INFORMAL FALLACIES

Pragmatics & Beyond

Companion Series
Editors:
Jacob L. Mey
(Odense University)

Herman Parret
(Belgian National Science Foundation,
Universities of Leuven and Antwerp)

Editorial Address:
Department of Germanic Languages
and Literatures
University of Antwerp (UIA)
Universiteitsplein 1
B-2610 Wilrijk
Belgium

Editorial Board:
Norbert Dittmar {Free University of Berlin)
David Holdcroft {University of Leeds)
Jerrold M. Sadock {University of Chicago)
Emanuel A. Schegloff {University of California at Los Angeles)
Daniel Vanderveken {University of Quebec at Trois-Rivières)
Teun A. van Dijk {University of Amsterdam)
Jef Verschueren {University of Antwerp)

Douglas N. Walton

Informal Fallacies
INFORMAL FALLACIES
Towards a Theory of Argument Criticisms

Douglas N. Walton
University of Winnipeg

JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA

1987
Library of Congress Cataloging in Publication Data
Walton, Douglas N.
Informal fallacies.
(Pragmatics & beyond companion series, ISSN 0920-3079; 4)
Bibliography: p.
Includes index.
1. Fallacies (Logic) 2. Logic. I. Title. II. Series.
BC175.W34 1987 165 86-27055
ISBN 90-272-5005-7 (Eur.)/l-55619-010-7 (U.S.)(alk.paper)
© Copyright 1987 - John Benjamins B.V.
No part of this book may be reproduced in any form, by print, photoprint, microfilm, or
any other means, without written permission from the publisher.
 ACKNOWLEDGEMENTS

I would like to thank Herman Parret for first suggesting the feasibility
of a project of this sort. Tom Beauchamp, Monroe Beardsley, and Henry
W. Johnstone, Jr. have helped by offering constructive criticisms of earlier
drafts of this manuscript. Among the many colleagues whose conversations
and written remarks have helped to clarify and improve my arguments I
would especially like to thank Charles Hamblin, Dick Epstein, John
Woods, Lynn Batten, David Harrah, Ruth Manor, Jim Mackenzie, Jaakko
Hintikka, Peter Miller, Krister Segerberg, Stan Surma, John Bishop, Jan
Crosthwaite, George Hughes, David Lewis, Peter Geach, Tony Blair, and
Ralph Johnson. Max Cresswell made some suggestions during a Logic
Seminar in March, 1983, at Victoria University of Wellington (New Zea­
land) that turned out to be both fundamental and fruitful in my approach to
the theory of logical dialogue-games. For sending materials, published or
unpublished, I would like to thank Nuel Belnap, David Harrah, Ruth
Manor, Jim Mackenzie, Erik Krabbe, Rainer Hegselmann, Dick Epstein,
Frans van Eemeren, and Rob Grootendorst.
This book was helped along considerably by research supported
through grants from the University of Winnipeg and the Social Sciences and
Humanities Research Council of Canada. I am grateful to Mrs. Amy Mer-
rett for production of the typescript. I would like to thank the editor of
Logique et Analyse for permission to reprint my article 'Enthymemes'
(Logique et Analyse 95-96, 1983, 395-410), included with some changes as
part of chapter 5. Finally my thanks to Harry Simpson for proof-reading the
manuscript.
 TABLE OF CONTENTS

CHAPTER 1: A NEW MODEL OF ARGUMENT 1

1. Introduction to the Fallacies 3
2. Some More Fallacies 9
3. Fallacies Combined in Realistic Dialogues 11
4. What is an Argument? 16
5. Criticism as Challenge and Response 22
6. Basic Categories of Argument Study 26
Notes: Chapter 1 31

CHAPTER 2: HOT RHETORIC AND ARGUMENT 33

1. Appeals to Popular Sentiment 33
2. Appeals to Force 39
3. Appeals to Pity 42
4. Overly Personal Argumentation 45
5. The Rhetorical Debate 48
6. Case Study: Parliamentary Debate 51
7. Conclusion 60
Notes: Chapter 2 61

CHAPTER 3: THE LOGIC OF PROPOSITIONS 63

1. Deductive Validity 63
2. Formal Logic 65
3. Classical Propositional Calculus 68
4. Applying Deductive Logic to Arguments 73
5. Invalidity and Fallaciousness 77
6. Relevance and Validity 79
7. Subject-Matter Relatedness 82
8. Relatedness Logic 84
9. Semantics and Pragmatics 88
10. What is a Fallacy? 93
Notes: Chapter 3 96
VIII TABLE OF CONTENTS

CHAPTER 4: LOGICAL DIALOGUE-GAMES 97

1. Different Approaches to Formal Dialogues 98
2. The Ad Ignorantiam Fallacy 105
3. Fallacies of Question-Asking 108
4. The Fallacy of Many Questions 110
5. Demanding Direct Answers to Questions 113
6. Misconception of Refutation 115
7. Case Studies of Political Debates 118
8. A Game with Dark-Side Commitments 125
Notes: Chapter 4 130

CHAPTER 5: ENTHYMEMES 133

1. The Tradition of Enthymemes 133
2. The Objectives of Dialogue 136
3. Veiled Commitment-Sets 142
4. Strategy and Plausibility 145
5. The Problem Resolved 146
6. Order of the Premisses 150
7. Multiple Premisses in Complex Arguments 154
Notes: Chapter 5 156

CHAPTER 6: LONGER SEQUENCES OF ARGUMENTATION 157

1. Sequences of Argumentation 158
2. Graphs of Arguments 160
3. Case Study: Argument on Sex Education 162
4. Case Study: Circular Argumentation 167
5. Plausibility Conditions on Arguments 170
6. The Missing Links 176
7. Conclusions on Circular Arguments 180
Notes: Chapter 6 183

CHAPTER 7: FALLACIOUS ARGUMENTS FROM AUTHORITY 185

1. How Appeals to Authority Can Go Wrong 185
2. Plausible Argument 188
3. Where Experts Disagree 190
4. Expertise and Legal Dialogue 192
5. Dialogue and Expertise 194
 TABLE OF CONTENTS IX

6. Conclusions 199
Notes: Chapter 7 201

CHAPTER 8: VARIOUS FALLACIES 203

1. Inductive Fallacies 203
2. Deductive and Inductive Arguments 205
3. Post Hoc Arguments 206
4. Slippery Slope 209
5. Equivocation 212
6. Amphiboly 213
7. Composition and Division 214

CHAPTER 9: ARGUMENTS AGAINST THE PERSON 217

1. Poisoning the Well 217
2. The Sportsman's Rejoinder 222
3. Evaluating Ad Hominem Disputations 227
4. Four Types of Circumstantial Ad Hominem 230
5. Rhetorical Context of Ad Hominem Attacks 233
6. Positional Defensibility 236
7. Conclusion 238
Notes: Chapter 9 240

CHAPTER 10: EQUIVOCATION 241

1. What is Equivocation? 242
2. Vagueness and Criticisms of Equivocality 245
3. The Problem of Subtle Equivocations 249
4. Deep Deception and Equivocal Dialogue 254
5. Many-Valued Logic for Equivocators 259
6. Priests's System LP 261
7. Applying LP to the Fallacy of Equivocation 266
8. R-Mingle as a Logic for Equivocators 273
9. RM and Equivocation 276
10. Conclusions 283
Notes: Chapter 10 287

CHAPTER 11: INFORMAL LOGIC AS A DISCIPLINE 289

1. The Role of Formal Logic 291
χ TABLE OF CONTENTS

2. Dialectic as a Theory of Argument 294

3. Function of Why-Questions 297
4. Subject-Specific Nature of Arguments 302
5. Case Studies on Circular Reasoning 305
6. Conversational Pragmatics 314
7. Pedagogical Directions for Informal Logic 320
Notes: Chapter 11 322

BIBLIOGRAPHY 323

INDEX 331
 CHAPTER 1: A NEW MODEL OF ARGUMENT

The basic question of this monograph is: how should we go about judg­
ing arguments to be reasonable or unreasonable? Our concern will be with
argument in a broad sense, with realistic arguments in natural language —
for example, persuasive arguments that might occur in advertising or in
political debates. Our concern will not extend to all aspects of such argu­
ments however. The basic object will be to engage in a normative study of
determining what factors, standards, or procedures should be adopted or
appealed to in evaluating an argument as "good," "not-so-good," "open to
criticism," "fallacious," and so forth. Hence our primary concern will be
with the problems of how to criticize an argument, and when a criticism is
reasonably justified.
To criticize an argument is, sometimes, to begin to take part in the
argument. In any such step of real engagement in the argument, there is a
danger of loss of the proper critical perspective, a danger of taking sides in
the argument itself. Can this step be avoided? Our problem is to show how
criticism can be reasonable. So we need to show how reasonable criticism
can itself contend with the criticism of having taken sides in the argument.
We need to pursue answers to two questions. First, what is an argu­
ment? And second, how is reasonable criticism of an argument possible?
Beyond these two questions, we will need to study and explore the kinds of
criticisms of arguments that have traditionally been thought to be reasona­
ble and to logically compel our consideration in judging arguments. Many
of these criticisms can be identified with the domain of the traditional infor­
mal fallacies, descended from Aristotle's sophistici elenchi, and long a part
of the logic curriculum. Much of our work will be in trying to make sense
out of these traditional fallacies as representing reasonable or unreasonable
types of criticisms that can be made of arguments. In that regard, our project
falls under the subject of applied logic, or informal logic as it often is called
these days. 1 In another way, our project could be looked at as a kind of
theoretical linguistic investigation in the field now called the pragmatics of
discourse.
2 INFORMAL FALLACIES

There are many requirements for a theory of argument. But in addi­

tion to the factors already mentioned, the following phenomena of argu­
ment must be dealt with. We list these matters briefly here to give the
reader some rough guidelines for what we take to be the proper subject
area of the theory of argument. First, an argument must be understood not
exclusively as a set of isolated propositions (a set of premisses and conclu­
sion), but also as an extended chain of arguments in the context of a con­
tinuous discourse, issue, or discussion. Second, the notion that each partic­
ipant in an argument may have a certain stance or position to defend must
be made sense of. Third, the phenomenon of a shift in the burden of proof2
in argument is often fundamental in correctly judging the force or value of
a criticism.
A fourth consideration is the distinction between different kinds of
argument. For example, it is often held that an argument may be induc­
tively strong even while it is deductively invalid. Does this mean that there
are two different kinds of argument, or merely that there are two different
standards for the strength of an argument? What other kinds or standards
of argument are there?
Fifth, we need to look at the borderline between arguments and non-
arguments. For example, overly personal attack and emotional diversion
are sometimes thought to be forms of non-argument and evasion of argu­
ment. Sixth, we need to understand the idea that some arguments are
defensive, in contrast to other arguments that attack a position not one's
own. The pro and con of argumentative discourse needs to be understood
as part of the argument itself.
Seventh and eighth, we need to look carefully at the role of formal
logic and rhetoric in the theory of argument.
A fundamental problem of applied logic is the fact that a good part of
most real arguments remains unstated by the arguer, yet is reasonably
taken as part of the argument by the one to whom the argument is addres­
sed. This is the traditional question of enthymematic premisses. An
enthymematic premiss is traditionally said to be a premiss not explicitly
stated in an argument, but tacitly presumed by the arguer. In a given argu­
ment, there may be one or more enthymematic premisses. It is generally
held that an enthymematic premiss is necessary for the argument in ques­
tion to be valid, and that all the required enthymematic premisses are suffi­
cient to make the argument valid. An enthymeme is an argument that has
one or more enthymematic premisses. We need to study the problem of
 A NEW MODEL OF ARGUMENT 3

defining the set of missing premisses that a critic may fairly be allowed to
add in to the argument he wishes to criticize. Enthymemes will be the sub­
ject of chapter 5.
The ninth area of concern is that of questions and answers. No theory
of argument could be adequate without dealing with the role that questions
play in arguments. Another aspect of argument that is important to deal
with concerns the various rules or procedures that can be adopted to regu­
late arguments. In specific contexts, arguers may agree on helpful pro­
cedural rules. For example, in legal contexts, certain forms of questions
may not be allowed to be asked. If Grice is right, there may even be pro­
cedural agreements — for example, an agreement to remain polite — that
are part of ordinary conversations.
An eleventh topic is the role of language in arguments. This topic
includes problems of vagueness, ambiguity, and definitions of terms. A
twelfth major topic is that of judgements of relevance of an argument or
criticism in relation to the issue or topic of the argument.
In this monograph we will argue for the pluralistic thesis that there are
many different models of argument, each of which may have a certain
legitimacy in relation to its context. However, the twelve requirements
above are so broad that we will concentrate on one model of argument as
paradigmatically central. We call this model of argument the dialectical
model, meaning that argument is thereby conceived as an interpersonal
exchange that takes place between two arguers. This model of argument
derives from Aristotle historically,3 and is quite a separate tradition from
Hegelian or Marxist use of this term to designate a historical process of
evolution. Rather, dialectic in our sense refers to a logical game of
dialogue, a verbal sequence of question-and-answer moves where the
objective of each player is to prove a thesis to the other.

1. Introduction to the Fallacies

There are fifteen (plus) or so major fallacies given by the standard
treatment of current and traditional logic texts, depending on how you
divide them up. 4 Each of these has traditionally been thought to be a seri­
ous and systematic type of error or fault in reasoning worth study in the
logic curriculum. The basic problem with the fallacies has always been that,
once you look at them closely, you can see that in each case the type of
argument alleged to be "fallacious" can evidently, in some instances, be a
correct — or at least not unreasonable — form of argument, or kind of
4 INFORMAL FALLACIES

move to make in argument. 5 In short, the would-be "fallacies" are not

always fallacious.
On the other hand, it remains clearly true that in some circumstances,
each, or at least many of these traditional fallacies can contain serious com­
mon forms of lapses of logical reasoning eminently worth systematic study.
And indeed, it will not take too much convincing to see that we deeply need
to understand the structure of each of the "fallacies" if we are ever to have
a useful grasp of how logic can be applied to realistic argument in a practi­
cal way. If there can be an applied logic, it must accommodate (at least) the
subject-matter of the major fallacies (along with our other twelve topics as
well).
With each of the fallacies, it is necessary for us to work towards under­
standing principles or procedures for helping us to adjudicate when the
"fallacy" really is an incorrect move in argument and when the alleged or
possible fallacy can be justified as a legitimate or justifiable move to have
made in argument. Thus each of the fallacies presents us with an analytical
problem. What is important to see, at the present stage of development of
applied logic, is that each of these problems can be solved. Solutions to
each of these problems are not immediately forthcoming in most cases, for
various reasons that will become clearer as we engage the problems. But a
shift towards a new conception of argument will at least provide a
framework in which each problem can be formulated so that it can be sol­
ved. This step toward a new theoretical conception of argument, initiated
most notably by Hamblin (1970), but now by others as well in various dis­
ciplines, brings out the advantages of looking at the fallacies (plus some
other matters) as related parts of a common underlying concern about the
theoretical modelling of dynamic argumentation.
Hamblin (1970) has already done a useful job of outlining the major
fallacies and indicating some basic questions about the status of each as an
identifiably incorrect type of argument. It is not helpful to cover this ground
again. But it may be helpful to the reader to briefly characterize the tradi­
tional or standard conceptions of each of the major informal fallacies —
including a few related topics — and show how each poses a particular
problem for analysis. In each case the basic problem is posed by the fact
that the "fallacy" in question is not, in every instance, a wrong argument.
The problem is: when is it right and when is it wrong?
One group of fallacies has to do with "hot" appeals to emotion in argu­
ment. Difficulties in reasonable argument and dialogue having to do with
 A NEW MODEL OF ARGUMENT 5

the illicit use of emotional appeals are perhaps the most common types of
problems in criticizing political debates and other forms of contestive argu­
ment in natural language.
The ad populum fallacy is the illicit use of "popular pieties" in argu­
ment. This fallacy is said to occur where appeal to popular feelings, or pre­
judices, is used to rouse conviction in an audience in lieu of presenting
genuinely relevant evidence. 6 Sometimes called "appeal to the gallery," this
fallacy can take place where an arguer zeroes in on powerful positive feel­
ings of a specific target audience and exploits these sentiments. Such feel­
ings may include patriotism, ethnic pride, or any special interests that con­
stitute a strong, positive pull of allegiance or emotional elicitation for the
person or group to whom the appeal is made. On the other hand, some­
times an ad populum appeal to negative feeling is made by invective against
those perceived as enemies by the audience. We will study several examples
of this fallacy in chapter 2.
The ad misericordiam argument is the appeal to pity, a familiar enough
strategy in argument. One must be careful here to note that appeals to pity
may, in some instances, not be unreasonable as a basis for making a deci­
sion. For example, charitable appeals for funds to support orphans may not
be advancing a fallacious argument in every instance where an appeal to
pity is made. But in some cases, an appeal to pity may be an evasion of rel­
evant considerations needed to make a decision on the issue. For example,
in a criminal trial, if the defence attorney bases his whole argument on an
appeal to pity, it could be quite reasonable to criticize his argument for its
failure to look at the evidence for the defendant's guilt or innocence. So
judgement is needed to evaluate whether, in a particular case, an argument
has made an illicit use of appeals to emotion.
Cederblom and Paulsen (1982, p.101) cite the case where a friend asks
you to write a letter of reference for a job he is not properly or suitably
qualified to undertake. Not wanting to hurt his feelings, you write the letter
saying he is qualified. Suppose your friend appealed to pity in his request
for you to write the letter. So moved, you disregarded your scruples, and
wrote it. Here your action could certainly be criticized. And if your only
reply was that you did it on the basis of pity for your friend, then your argu­
ment would be a weak one, and open to the ad misericordiam type of criti­
cism.
A third fallacy of emotional appeal is the ad baculum, or appeal to
force. Making a threat of force, or responding to one, is not in itself illogi-
6 INFORMAL FALLACIES

cal. But where the appeal to force is offered or taken as a reason for infer­
ring that a certain conclusion or recommendation is true, there may be an
ad baculum fallacy committed. As with the previous pair of fallacies, the
problem is to sort out, in a particular case, just when the emotional appeal
has become a fallacious argument.
A fourth fallacy of hot rhetoric is the ad hominem or ad personam fal­
lacy, where an argument has become an excessively personal attack on an
arguer's position. In the abusive ad hominem attack, the argument has
degenerated into personal abuse and vilification. In the circumstantial ad
hominem attack, it is alleged that the individual's personal circumstances or
actions are inconsistent with his arguments or policies. For example, a
politician who advocates wage restraint may be personally criticized for ask­
ing for an increment to his already high salary. This form of personal attack
of an arguer's circumstances can be quite legitimate. 7 But because it is a
personal and emotional form of argument, it can be carried too far in many
cases. Indeed, the ad hominem attack is such a powerful type of argument
that, in political and other adversarial debates, it is often accepted quickly
and uncritically by audiences as a knock-down refutation of a person's argu­
ment. Yet characteristically, it is quite a complex matter to sort out, in a
given instance, to what extent and in what ways an ad hominem criticism is
reasonable or fallacious. We will look at an example of ad hominem attack
in section 5 of this chapter.
A fifth traditional fallacy associated with appeals to personal credibility
is the ad verecundiam, or appeal to authority. Appeals to the authority of
expertise on a topic can be quite legitimate, and are often so taken by the
textbooks. But clearly, such appeals can fail in various ways.8 The "expert"
cited may not be an expert at all in the field at issue, for example. And
appeal to expert sayso may not, in every instance, be a good substitute for
looking at objective evidence, if it is available. Hence appeals to authority
can be tricky and complex arguments to sort out and reasonably judge.
Of the fallacies categorized by Aristotle as being dependent on lan­
guage, four will turn out to be important for our purposes. Equivocation is
the fallacy that arises through the use of ambiguous terms in an argument.
Consider the following argument.
Partisans are not to be trusted.
Democrats are partisans.
Therefore, democrats are not to be trusted.
 A NEW MODEL OF ARGUMENT 7

What seems worth noting here is that a shift has occurred. 'Partisans' means
'people who are biased and prejudiced' in the first premiss. But the same
word would seem to be used in a different way in the second premiss,
meaning 'members of a political party'. Presuming that there is a significant
difference between these two meanings, the argument above fails to estab­
lish that its conclusion is true, even if it is granted that both premisses are
true. Reason: even though the argument may appear on the surface to be
valid, it is likely that most persons, to whom the argument would be
directed, would take one premiss the one way, and the other premiss the
other way. However, so interpreted, the sentences of the premisses do not
go together as propositions to make up a valid argument for the conclusion.
Hence equivocation can be fallacious.9
Equivocation can have to do with vagueness as well as ambiguity. For
terms in natural language, because they are intrinsically vague, may be
open to varying disambiguations. Consider the following argument.
An elephant is an animal.
A grey elephant is a grey animal.
Therefore, a small elephant is a small animal.
Here we have a relative term, 'small', that shifts meaning according to the
context. A small house may not be taken, in some contexts, as anywhere
near the size of a small insect. 'Small' is a highly relative term, unlike 'grey',
that shifts according to subject. A small elephant is still a relatively large
animal.
The notorious slippery slope fallacy arises out of the vagueness of
terms in natural language. Each time you remove a grain of sand from a
heap of sand, it still remains a heap of sand. But if you keep applying that
principle, each time taking away one grain, then eventually you will no
longer have a heap. Yet there is no one single point you can say where the
principle has failed — there is no single step where the heap ceased to
become a heap because one grain of sand was taken away.10
So in the same way, a student might argue that although 50 is the low­
est passing grade on a test, if he got 49, he should be passed. For he might
argue: it is absurd to suppose that the one mark between 49 and 50 is educa­
tionally significant enough to withhold someone from passing for the lack of
that one point. But if the appeal were granted, a student who received a
mark of 48 could argue, with equal justification, that he should receive 49,
and therefore also pass the exam. Such an appeal could be repeated, with
8 INFORMAL FALLACIES

equal justification, until no one student could receive a failing grade.

Slippery slope arguments may be difficult to repel if a border or
guideline is perceived as being somewhat arbitrary. Such an argument calls
for a recognition of the vagueness in the standard on which the guideline is
based. But once conceded, a vagueness without precise limits — remember­
ing that any criterion with precise limits is no longer vague — may be open
to attacks that chip away at the guideline, one step at a time.
Amphiboly, meaning 'double arrangement' is the same fallacy as
equivocation except that the double meaning involves a group of terms in a
grammatical construction where no single term in the group is ambiguous in
itself. An example of this sort of grammatical double meaning is given by
Engel (1976, p.52): "With her enormous nose aimed at the sky, my mother
rushed toward the plane". Although this sentence certainly illustrates a
grammatical incongruity, it is not an argument, by itself. Hence it is ques­
tionable whether the sentence represents a fallacy of amphiboly.
Amphiboly, although it has been traditionally classified as a significant
fallacy by textbooks, presents several problems. If it is truly a fallacy, then
we need to see how grammatical double meaning somehow combines with
argument to produce a fallacy. The usual examples of the texts, like the one
cited above, fail to suggest how amphiboly could be a serious fallacy in its
own right. However, amphiboly seems to be a special, and perhaps
unusual, case of equivocation, and we will see that as a fallacy it could share
many problems of analysis with equivocation.
The final fallacy of this group we will mention is the dual fallacy of
composition and division. This fallacy has to do with arguing from the prop­
erties of a part to the properties of a whole or vice versa. First, let's take an
example of composition. Because all the members of a football team are
good, for example, it does not follow that the team is good. Or to take an
example of division, because a machine is heavy, it does not follow that a
particular part of the machine is heavy.11
Traditionally, authors have attempted to give various schemes of clas­
sifications of the fallacies. However, as Hamblin (1970) has shown, these
schemes are nothing more than ad hoc devices of little serious import. We
attempt no scheme of this sort. But for the reader's convenience, it may be
helpful to group the fallacies of hot rhetoric and the fallacies dependent on
language as two roughly grouped areas to begin with in our presentation.
The remaining fallacies are even more heterogeneous however, and we will
not try to group them rigorously in any special order or scheme of classifica-
 A NEW MODEL OF ARGUMENT 9

tion. Instead, we categorize seven basic faults of argument in section 6 of

this chapter. But by the end of the monograph the reader will see that the
domain of the fallacies is far too complex to classify easily.

2. Some More Fallacies

The ad ignorantiam argument, or argument from ignorance, occurs
where lack of knowledge is used as a premiss to argue to the conclusion that
some proposition is true or false. It is said to be one form of this fallacy to
conclude that a proposition is false merely on the grounds that the truth of
the proposition has never been established. Alternatively, it is said to be a
second form of this fallacy to conclude that a proposition must be true, sim­
ply because it has never been disproven.12
The stock example is the argument that ghosts do not exist because
nobody has ever been able to prove that ghosts do exist. While it is easy to
appreciate the sort of mistake at issue, one must be careful, because some­
times arguing from ignorance is not altogether unreasonable. If you are not
sure whether eating a "mushroom" you have just picked is safe, better to at
least provisionally assume that it is not safe. Here the negative conclusion
may be reasonably justified by lack of knowledge only.
The fallacy of ρetitio principii, sometimes also called begging the ques­
tion or arguing in a circle, is said to occur where a conclusion is "proved" by
assuming the conclusion as one of the premisses of your argument. For
example, suppose a philosopher wants to prove that induction is justified as
a reliable source of knowledge, and he sets out to prove it by giving several
examples, where, he suggests, induction works. A critic might argue that
this philosopher has begged the question because he has presupposed the
reliability of induction, presuming that the philosopher had argued induc­
tively on the basis of the examples he used. 13
The problem with arguing in a circle as a fallacy is that it's not always
clear that circular arguments are fallacious. Suppose an economist is asked
why the economy in a certain province x, is in a slump at the present time,
and he offers a number of reasons, including the fact that many people have
been leaving province x. But suppose further that while giving reasons why
people are leaving the province, the economist cites the relative prosperity
of other provinces, in effect making the presumption that the economy is in
a slump in province x. The economist's argument is circular, when taken as
a whole. But is the circle necessarily fallacious? The economist could con­
ceivably defend the circularity of his argument by pointing out that there is
10 INFORMAL FALLACIES

in fact a circular feedback relationship between the behavior of people leav­

ing province χ and their correct perception that the economy is in a slump
in that province. If so, the circularity in his argument need not necessarily
be fallacious in the vicious way.14
In this case and others, if a critic perceives a circle in someone's argu­
ment, it may be a reasonable criticism to point out the circle. But in some
cases, the arguer might be able to respond to the criticism by showing that
the circle does not necessarily undermine or refute his argument altogether.
Hence the general problem of the petitio principii is to distinguish between
circles that are vicious and those that are benign, or at least can be dealt
with or defended.
Another group of fallacies has to do with the asking of questions. The
fallacy of complex question is illustrated by the famous example "Have you
stopped beating your spouse?" This type of question suffers from many
faults, not least that it unfairly attempts to force an answerer to concede the
prejudicial presumption of being a spouse-beater — no matter which way
he answers, 'yes' or 'no'.
Another allegation of fallaciousness pertaining to questions occurs
when an answerer fails to give a direct answer to a question, yet gives an
answer that is indirectly related to the question. Such a response may be
dismissed by a critic as irrelevant, as a red herring. This brings us to the
topic of relevance in argument. 15
All the fallacies discussed so far could be, and have been, called falla­
cious on the ground that they are failures of relevance in argument. Yet
some texts follow the Aristotelian tradition of singling out one term for the
special fallacy of failure of relevance.
The traditional term ignoratio elenchi, meaning "ignorance of refuta­
tion" is often translated as "irrelevant conclusion." According to Hamblin
(1970, p.31), Aristotle meant this term to refer to the case where someone
offers an argument that may even be valid but has the wrong conclusion. In
other words, the fault appears to be that the arguer thinks he has proved
one thing, but has really proved something else.
The straw man fallacy, on the other hand, is the mistake of getting the
premisses of an argument wrong. Here, the fault is that of incorrectly or
inaccurately attributing a position to an arguer that he does not really
accept, or has given evidence of having accepted. This is indeed a common
fault of argument, or at any rate, arguers quite often try to criticize their
opponents for having misrepresented their views or their position. And of
 A NEW MODEL OF ARGUMENT 11

course, such criticisms are sometimes justifiable. Here the problem is that
although an argument may be quite valid, it may commit a straw man fal­
lacy if drawn from premisses that do not fairly represent the position it is
meant to refute.
Another group of fallacies are the inductive fallacies. The most notori­
ous of these is the post hoc, ergo propter hoc fallacy,16 where a correlation
between two events A and  is fallaciously used to draw the conclusion that
A causes B. The chief problem here is that there appears to be little firm
agreement on how the relationship of causality is to be precisely defined.

3. Fallacies Combined in Realistic Dialogues

In real arguments, moves open to critical analysis and questioning may
occur at various junctures in a sequence of dialogue where each juncture is
related to one of the traditional informal fallacies. In this fashion, several of
the fallacies may be combined in a given passage of dialogue. However,
typically, each "fallacy" is better described as a move or allegation by a
party to the argument that could be open to reasonable questioning, rebut­
tal, or critical questioning by the other party, or by some third-party critic.
In the present case, these third-party critics are you, the reader, and myself,
the author of observations on the passage of dialogue being considered.
Consider the following sequence of dialogue from Hansard {Canada:
House of Commons Debates, Feb. 17,1984, p. 1499),17 The topic of this par­
liamentary debate was tax proposals affecting professionals.
Hon. Edward Broadbent (Oshawa): Mr. Speaker, I would like to move on
to another aspect of the unfairness as it pertains to the possibilities open to
professional people such as doctors and lawyers. The Budget provision is
that doctors, lawyers, and other professional people who may be earning
$80,000 to $90,000 a year can now re-establish themselves as companies.
Does the Prime Minister consider it fair that professionals and other upper
income people who are paying an effective tax rate of 50 per cent now
because their income is in the $80,000 to $90,000 category, should be able
to have their taxes reduced to about 15 per cent, which is the level an ordi­
nary worker has to pay who has not open to himself the same privilege of
reconstituting himself as a company? Is this the kind of fairness the Gov­
ernment believes in, that there should be a special hand-out in terms of
redistribution of income to richer Canadians instead of moving toward jus­
tice for middle and low income Canadians?
Right Hon. P.E. Trudeau {Prime Minister): Mr. Speaker, the Minister of
Finance is proposing a way of treating professionals and other such people
12 INFORMAL FALLACIES

which has been used in the past, and to which he is returning now. We had
another system for a period of time. I do not recall the Leader of the NDP
having supported it in any vocal way.
I do recall when the former Minister of Finance, the present Secretary of
State for External Affairs, had a budget in November, 1981, which was
deliberately attempting to close what the Hon. Member has called
loopholes, that he was not supported by that Leader of the NDP, nor by
his colleagues. It is quite typical of the NDP that they will not support any
progressive reform by the Government. They will merely continue to
attack everything indiscriminately.
Mr. Broadbent: It is quite typical of the political style of the Prime Minis­
ter to shift in a sleazy way from a difficult question that he was supposed to
answer, and provide irrelevant and misleading information. This Party
supported the provisions in that Budget which would remove the loopholes
that go to upper income Canadians. We opposed the other provisions that
gave other benefits to people who did not deserve to get them. That should
be clear.
If you look over this sequence of question, reply, and objection, you see the
following moves that can be identified with traditional fallacies. First, Mr.
Broadbent asks two very complex questions. Both questions could be
described as loaded and prejudicial. In reply, Mr. Trudeau launches a cir­
cumstantial ad hominem attack against Mr. Broadbent. Finally, Mr. Broad­
bent accuses Mr. Trudeau's answer of being a "sleazy shift" away from
answering the question. In effect, Mr. Broadbent accuses Mr. Trudeau of
having committed a fallacy of irrelevance.
The example is an interesting one for many reasons. But it is especially
instructive to observe how fallacies of question-asking, ad hominem, and
irrelevance can all be combined in this one fairly brief but stormy inter­
change in debate. What can be said about these matters at this point?
First, let us examine Mr. Broadbent's questions. The first question is
whether Mr. Trudeau considers it fair that a professional should be able to
reduce his taxes by classing himself as a corporation. This question does not
seem unreasonable, if Mr. Broadbent's figures and assumptions about the
present tax regulations are correct. The question is a lengthy and highly
complex one, and there are some aspects of it that could well be criticized
by an answerer. But the next question is clearly more open to criticism, so
let us pass on to it.
Mr. Broadbent describes the present tax policy of the Liberal govern­
ment to that party's leader, in this question, as a "special hand-out" to
 A NEW MODEL OF ARGUMENT 13

"richer Canadians". In effect, he is suggesting that this policy is not fair, as

he makes clear by going on to say that the policy is one that gives a hand­
out to the richer Canadians "instead of moving toward justice for middle
and low income Canadians". Of all this — Broadbent queries — is this the
kind of fairness the government believes in? Clearly the question is ironic.
What Mr. Broadbent is really asking is whether that is the kind of unfair­
ness that the government believes in.
The grammatical form of this question is that of the yes-no question.
That is, there are only two options for giving a direct answer to the ques­
tion: 'yes' or 'no'. If Mr. Trudeau answers 'yes', then what he thereby con­
cedes is that he believes in the kind of "unfairness" described by Mr.
Broadbent as giving hand-outs to the rich and depriving others (the non-
rich) of justice. On the other hand, if Mr. Trudeau answers 'no', he thereby
concedes that he does not believe in the official government tax policies, set
by the Liberal Party, the party Mr. Trudeau leads. Clearly, if he were to
answer 'no', Mr. Trudeau would immediately be attacked by Mr. Broad­
bent on the ad hominem accusation of being inconsistent with the position
of his [Mr. Trudeau's] own party. In short then, no matter which way Mr.
Trudeau answers, he commits himself to an extremely damaging admission
of guilt or inconsistency.
But the question is a yes-no question. To answer it directly, Mr.
Trudeau is offered only these two options — yes or no. Is it a fair question?
Clearly it is not. It is a complex question and a loaded question. Mr.
Trudeau would be foolish indeed to give a direct answer to it. He should, in
all fairness, be allowed to question some of the presuppositions inherent in
the question. He should, in short, reasonably be allowed to avoid giving a
direct answer, without altogether evading the issues raised by the question.
Here then we are confronted with the problem of the many questions
fallacy and associated problems with the asking and answering of questions.
Asking a question can be fallacious. And failure to answer it directly
enough, given the circumstances, can also be fallacious.
Mr. Trudeau answers the question by directing an ad hominem attack
against Mr. Broadbent's position on the issue of taxation of professionals.
He replies that in the past, professionals could not use these means of
reducing their taxes. Mr. Trudeau then claims that he cannot recall Mr.
Broadbent's ever having supported that previous policy in a vocal way. So
Mr. Trudeau alleges, in effect, that Mr. Broadbent is inconsistent. He
never supported the old policy in a vocal way, but now he vocally supports
14 INFORMAL FALLACIES

it. Why? According to Mr. Trudeau's reply, it is because Mr. Broadbent

and his NDP colleagues simply attack all government policies indiscrimi­
nately. Mr. Trudeau claims, in effect, that Mr. Broadbent's attack on tax
policies is not based on a consistent or reasonable position, but merely on a
typical NDP reflex attack on all government policies. This rejoinder is, in
effect, an attack on NDP integrity and policymaking generally.
Now the question — has Mr. Trudeau unfairly avoided answering Mr.
Broadbent's question? Has he committed a fallacy of irrelevant answer?
First, we need to note that Mr. Trudeau's statements in answer are, in some
sense at least, relevant to the question. They do relate to tax policies. Sec­
ond, we have already seen that Mr. Broadbent's question, as put in its pre­
sent form, was not a fair or reasonable question to answer directly. Third,
although Mr. Trudeau's answer is an ad hominem in the sense that it attacks
Mr. Broadbent's personal voting record in relation to his present argument,
and criticizes the consistency of the NDP position, it may be that not all ad
hominem arguments are fallacious. In some situations, pointing out that an
arguer does not practice — say in his voting record — what he now
preaches, could be a reasonable form of criticism in argument.
For these reasons, Mr. Broadbent's accusation in reply that Mr.
Trudeau has shifted "in a sleazy way from a difficult question" is at once
ingenuous and exaggerated. The question was more than difficult — it was
impossible to answer directly! And to call the reply "sleazy" is to use an
emotional and inappropriate word to describe the ad hominem attack.
Given the unfairness of the phrasing of the question, one is almost
tempted to think that Mr. Trudeau's ad hominem reply is a fair enough
answer. But that is not the point, at least entirely. We should ask whether
the point was evaded by the ad hominem reply. Should Mr. Trudeau have
offered further positive support for the government tax policy on profes­
sionals instead of attacking the NDP consistency on this issue? Perhaps he
could have. On the other hand, his query of the opposing party's consis­
tency on this topic is not essentially unreasonable as a form of argument.
Mr. Broadbent does reply to Mr. Trudeau's attack by claiming that his
party has been consistent. He claimed that his party had supported some
aspect of the past tax policies, and only opposed the new provisions. How­
ever, he now describes these provisions as "loopholes that go to upper
income Canadians." Once again, the language is excessively loaded and
prejudicial. Mr. Broadbent is aggressively trying to beg the disputed ques­
tion of whether these provisions are good or fair aspects of tax policy. His
 A NEW MODEL OF ARGUMENT 15

reply then, is based on a sound point, but can be criticized for its excesses.
This excessive aggressiveness in attempting to discredit the opposing
party's position is quite typical of political debate as a form of argument.
This particular interchange has consisted in a series of parries, or attacks on
each other's position. While such an ad hominem approach is to some
extent legitimate within the adversarial structure of parliamentary
debates, 18 one might also wonder whether excessive use of positional attack
represents a lower grade of dialogue than one should expect in parliament.
Clearly, Mr. Trudeau had to respond aggressively to support his
party's position in face of Mr. Broadbent's nasty question. Yet one could
question whether Mr. Trudeau's method of refutation of the question — by
claiming that the NDP typically attacks the government — could be an
over-reactive, unsupported generalization. Surely it is not a highly plausible
thesis that the NDP attacks the government indiscriminately as a quite gen­
eral practice in their arguments. Even if it is true, it would be a difficult
general claim to establish, and Mr. Trudeau would have been on easier
grounds to support if he confined himself to the present topic of tax reform
in the case of professionals. Or perhaps he could have cited a particular
case where the NDP attacked a government policy that subsequently
showed evidence of success. Therefore, even in light of the adversarial
nature of partisan debate, one could question whether the use of ad
hominem argumentation is in this case excessive.
We see then that just to say that one party committed a "complex ques­
tion" fallacy, or that the other committed an ad hominem fallacy is by no
means the final word. Each of these so-called fallacies is a form of argument
that is part of a certain context of dialogue. In light of the objectives of the
dialogue then, each move may be critically questioned and evaluated as
being reasonable in some respects, but open to criticism in others. One
criticism may be best evaluated in relation to the previous move in dialogue
that it was made in response to.
In chapter 8, we will devote a whole chapter to the systematic analysis
of the ad hominem fallacy. For the moment, the lesson is that the reasona­
ble and fair evaluation of ad hominem argumentation is a complex matter,
and involves analysis of the context of dialogue in which the argument
occurred. That lesson will turn out to be true of all the fallacies and other
aspects of argument so far mentioned. Looking at an argument as simply a
designated set of propositions, the premisses and conclusion, does not take
us far enough to be much help.
16 INFORMAL FALLACIES

Hence we need to inquire more carefully into the nature of argument,

and introduce the idea that an argument is a form of dialogue interchange
between participants in the argument. Otherwise, there can be no hope of
evaluating the cut and thrust of realistic disputation as consisting of fair or
unreasonable criticisms.

4. What is an Argument?
The basic definition of an argument that has usually been given in logic
textbooks is that an argument is a set of propositions, one of which is desig­
nated as the conclusion. The remainder are called premisses. To determine
which is the conclusion, the reader usually gets different advice. But it is
often suggested that the conclusion is preceded by a word like 'therefore' or
that the premisses provide the "basis" for the conclusion. What this type of
suggestion amounts to, in practice, is not too clear. For very often the con­
clusion is not preceded by 'therefore' or some clearly equivalent term. And
what 'basis' means is no more clear than what 'premiss' is supposed to
mean. However, it has seemed satisfactory, for the purposes of such
textbooks, not to pursue such matters further and to lean heavily on the
word 'designated'.
When it comes to analyzing the fallacies however, that approach is not
good enough. The fallacy of ignoratio elenchi involves an arguer getting
wrong which is the conclusion for an argument. The straw man fallacy
involves wrongly identifying the premisses. Petitio principii may involve
confusing the conclusion with a premiss. Emotional fallacies like ad
populum sometimes involve the mistake of thinking there are premisses
when it is not clear that there are any propositions that can be clearly iden­
tified as premisses. All these fallacies, and others as well, involve a failure
to clearly or correctly identify or distinguish between the premisses and
conclusion of an argument or putative argument. Here, the definition of an
argument as a set of propositions is best seen as a necessary condition only
— a step towards a fuller account.
An argument is basically a set of propositions. But that is not all there
is to it. There is another factor. An argument is a claim made that the con­
clusion follows from premisses advanced. Therefore, an argument may be
defined as a set of propositions where one is claimed to follow from the
others.
Sometimes it is also suggested to be part of the definition of argument
that the premisses may serve to provide support for, or may be taken as the
 A NEW MODEL OF ARGUMENT 17

basis of the truth of the conclusion. For example, Copi (1982, p.6) writes
that an argument is "any group of propositions of which one is claimed to
follow from the others, which are regarded as providing support or grounds
for the truth of that one." This account of the concept of an argument
agrees nicely with the one proposed above except for the addition of the
clause beginning "which are regarded as ..." Is such a clause necessary or
useful? Some care is needed in discussing this question.
Initially, the point should be made that, in some arguments, the pre­
misses are meant to support "positive" or "direct" arguments. But in other
arguments, the premisses advanced are meant to criticize an opponent's
position, or to rebut objections to the conclusion meant to be defended. We
could call these "negative" or "indirect" arguments. Such arguments may
provide indirect support for a conclusion. However, it is important not to
confuse the difference between support for the truth of a proposition and
refutation of an objection that supports the falsehood of a proposition. For
it may be quite consistent to argue that there is no support for either the
truth or the falsity of a proposition. At least in some cases, if there is no rel­
evant evidence and no burden of proof, it can be reasonable to be a skeptic.
Hence there is reason to believe that the final clause may not be
characteristic of all arguments. Whether it is remains open to argument.
But let us get back to the idea that an argument is a claim for its con­
clusion. What could this mean? It seems to mean that there is a certain
thrust or direction or force in argument that goes from the premisses to the
conclusion. But such a notion of "thrust" is highly metaphorical. It is not
clear what it amounts to, in precise terms.
Many of the fallacies presuppose that there can be two conclusions in
an argument — the conclusion of the argument being criticized or evaluated
and the conclusion of the argument of the critic or evaluator. What emerges
is an interpersonal conception of argument. If you and I are engaged in
argument then presumably you may have your conclusion and I may have
mine. The two conclusions may be quite different, even logically opposed
in some cases.
For example, the concept of begging the question seems to presuppose
a context where each of two arguers has his own conclusion to be proven.
In order to prove my conclusion, I should "beg for it" from you. Rather
than simply asking you to grant my conclusion — in effect arguing 'C,
therefore C' — I should present something different from my conclusion 
as evidence you might be inclined to accept.
18 INFORMAL FALLACIES

In order for there to be an argument, there should be an "issue" to be

disputed or argued about. At least the informal fallacies strongly require
this notion. What this amounts to is that each party to the argument —
there may be more than one party — should have his or her conclusion for­
mulated clearly, if the argument is to have any point or purpose. Once my
conclusion in the argument is clearly formulated and established as such, I
may be fairly accused of ignoratio elenchi if I prove some other proposition
judged to have no real bearing on that conclusion.
In adopting the idea that there could be two conclusions in an argu­
ment however, clearly we are moving to a much richer conception of argu­
ment than the initial one of a set of propositions. We are moving toward the
first basic idea of logical dialogue, that each party to the argument should
have a thesis (conclusion) to be proven.
Sometimes in real-life arguments it is not established what the thesis of
each participant really is — the thesis that he or she is supposed to be argu­
ing for. In many quarrels, for example domestic quarrels and political
debates, the two parties are disputing and arguing, but it is not clear what
the real issue is.
This phenomenon means that logical games of dialogue are one step
removed from the realities of actual argumentation. In the games of
dialogue, the thesis of each participant that he is ultimately supposed to
prove must be clearly set as a proposition in advance of any play of the
game.
Many quarrels in real life are simply vague in this regard. The disput­
ants may never make clear precisely what they are quarrelling about. The
issue is not defined precisely.
It is interesting to reflect however that, in some instances, the dispute
may serve to help define the issue. A domestic quarrel may arise in heated
dispute over who is to take the garbage out. However, it may arise through
the course of the argument that the real issue is the husband's coming home
late the previous night. In such a case, the preliminary argument may have
served the valuable function of defining the real issue. Disputing about the
ostensible issue served as a process from which the real issue of contention
could emerge.
Once the issue is formulated and argument is underway, we might now
wonder what the purpose of an argument is. Very often, in real arguments,
the purpose of an arguer is to defeat the opponent at all costs. But if logical
dialogue is to represent a normative model of reasonable argument of
 A NEW MODEL OF ARGUMENT 19

interest in analysis of the fallacies, the means of winning should reflect

some sort of "logical process" of proceeding from premisses to conclusions.
This leads us to reflect on the conditions under which arguments are won or
lost when the participants are perceived as being reasonable in their
method of argument.
There are two basic ways to win an argument. The more familiar way
is to present your own arguments. This way means presenting your own
reasons for accepting the proposition at issue, in the hope that your listener
will also find these propositions as plausible and attractive as you do.
The second way is to listen to your opponent and try to argue from her
side of the fence. This way means looking at her reasons for persisting in
the opposite viewpoint, and trying to argue on that basis. There are several
ways of doing this. One is to take the opponent's commitments, and show
that they imply your own thesis that you want to persuade her is true.
Another is to show that your opponent's position is internally inconsistent
or questionable. Reductio ad absurdum is a type of argument of this second
form.
These two ways of winning an argument do not exhaust all the possible
avenues. A third way is to show that your reasons for accepting your thesis
are better than your opponent's reasons for not accepting it. A fourth way
is to show that your thesis follows from a set of propositions that both you
and your opponent accept.
The third and fourth ways involve the commitment-sets of both partic­
ipants in the argument. The first two ways, however, are opposed to each
other in one important respect. The first way argues from premisses in your
own commitment-set. The second argues from premisses exclusively in your
opponent's commitment-set.
There may be many purposes of argument. One might be communica­
tion or exchange of information.19 However, in analysis of the informal fal­
lacies we see that a characteristic objective is for one participant to per­
suade another that a certain proposition should be accepted. Such a "game
of persuasion" basically involves the following type of strategy — you try to
get your opponent to accept your conclusion on the basis of his premisses
(propositions he is committed to or has accepted previously). The object is
to use logical rules of inference to prove your conclusion from his premis­
ses. In this framework, a compelling argument is an argument that should
persuade a rational arguer to accept a conclusion. What one arguer needs
to provide a compelling argument is a set of propositions that are commit-
20 INFORMAL FALLACIES

ments of the person to whom the argument is addressed, where this set of
propositions implies the conclusion by rules that both parties accept.
A compelling argument may be a long process requiring a strategy
comprising several sub-arguments. Therefore in practice, many arguments
are more like promissory notes. They make a start towards achieving an
argument that will hopefully be compelling once it is constructed in full.
Hence many arguments will not be compelling as they stand. They merely
"lead in a certain direction." Our concept of argument, therefore, must be
broad enough to allow for chains of arguments made up of component
arguments.
The central function of argument as interpersonal persuasion then
involves an interchange of premisses and conclusions. This represents the
model of argument we might call dialectical because it truly represents
"dialogue" in the sense of systematically arguing from the other's premis­
ses. It is this art that presumably lies at the bottom of the expressed value
often placed on the humanistic empathy of appreciating the other person's
point of view. This concept of argument represents a model of reasonable
argument that rejects the dogmatic stance as a failure of reasonable argu­
ment.

my
premisses

my
conclusion

The dialectical model of argument is not the only legitimate model of

reasonable argument, but it is clearly central in attempting to understand
the fallacies as reasonable or unreasonable moves of criticism, argument
and refutation. However, there are many ways in which we might define
"win" and "loss" in such a dialectical game. We could have relatively loose
procedural rules for what counts as a win or loss. Perhaps we could simply
require that the participants be sincere and honest. But since the fallacies
are ostensibly about certain types of deception or subtle unfairness in
 A NEW MODEL OF ARGUMENT 21

moves of dialogue, it could be useful and interesting to construct precisely

regulated games (like chess or monopoly) where permissible moves lead to
a definite 'win', 'lose' or 'draw'. We will study both these options.
A final characteristic of argument as dialogue is that the basic moves
should consist of a sequence of questions and answers. In the most simple
type of exchange, the questioner might ask 'Do you accept proposition A?'
and when the answerer replies 'yes', proposition A is entered in his log of
commitments. In a symmetrical game, the participants take turns asking
questions.
The foregoing considerations all suggest that we adopt a new model of
argument where the main outlines are given by the following sequence of
definitions. Each of these definitions represents a rough beginning of the
theory to come. Each of them will be given alternatives and more precise
implementations in the sequel.
An argument is a set of propositions advanced by a player (the propo­
nent) in a game of dialogue. The conclusion is one proposition in this set,
designated by the proponent in accord with his strategic objectives. The
remainder of this set of propositions are called premisses. These proposi­
tions are selected by the proponent with the strategic objective that the
other player (the respondent) is either committed to them, or will become
committed to them during the course of the game. This basic conception of
what an argument is leads on naturally to the following conception of a log­
ical game of dialogue.
A game of dialogue is composed of two players, called the proponent
and the respondent, a set of moves of the players, and a set of rules for the
game. Each move is either a question or an answer. The players take turns
making moves. There are different kinds of procedural rules which deter­
mine when various kinds of moves are allowed in the game. There are logi­
cal rules, defining what counts as a valid argument. The player wins who
first proves his thesis from the other player's commitments by means of the
rules. There are rules of commitment, defining how commitments are
incurred by a player as a result of moves he makes. Each player starts the
game with a set of initial commitments, and the commitment rules add to
(or in some cases subtract from) the members of this set. The thesis that
each player must prove is set at the outset of the game. Each rule of the
game is also designated at the outset. A player may formulate strategies to
guide him in making moves that will enable him to prove his thesis from the
other player's commitments. The set of commitments of a player is called
22 INFORMAL FALLACIES

that player's position.

Using this basic conception of argument, we can now go on to examine
the individual fallacies, and try to see in each case what is the best system
of dialogue-rules for fairly managing and adjudicating on allegiations of fal­
laciousness. Alleged fallacies are sometimes defensible arguments, so we
want to search out rules to fairly regulate disputable cases.

5. Criticism as Challenge and Response

Dialectic is essentially a reactive form of argument in the sense that
every move made by an arguer needs to be seen as a reply or counter-move
to some move of the other party to the argument. For this reason, many
criticisms and other forms of argument have as their essential and legiti­
mate function what may be described as "putting the ball in the opponent's
court." For example, the legitimate function of the tu quoque form of the
circumstantial ad hominem is to take the opponent's criticism of your posi­
tion and deflect the very same criticism back onto his original attacking
criticism. The proper effect of the tu quoque rejoinder then is to deflect the
burden of proof away from your own position and back onto the opponent's
argument. Your opponent must, if the tu quoque is successful, then turn at
his next move to defending his own position instead of attacking yours.
Within the context of dialectic as a model of argument, the tu quoque
can be a reasonable form of argument. Well deployed, it can be an excel­
lent strategy in dialogue, to be sure. However, it can in some circumstances
be used fallaciously as a means of attempting to evade argument, or of try­
ing to avoid answering a reasonable question by setting up a diversionary
attack. Where this possibility exists, care must be taken in setting out to
evaluate whether the ad hominem rejoinder should fairly be judged permis­
sible or fallacious.
A real example of the tu quoque rejoinder will illustrate some finer
points to be considered in attempting to evaluate this kind of move in argu­
ment. The following dialogue is taken from Hansard (Canada: House of
Commons Debates, vol. 127, no. 43,1984, pp.1812-1813).20 The topic of the
debate was economic development, and in particular, a special capital
recovery projects program where employees determine priorities. The
especially interesting part of the debate concerns a question posed by Mr.
Thomas Siddon, and the reply given by Mr. Donald Johnston of the Liberal
Party (the party then in power).
 A NEW MODEL OF ARGUMENT 23

Mr. Thomas Siddon {Richmond-South Delta): Mr. Speaker, my question is

directed to the Minister of State for Economic and Regional Develop­
ment. It refers to the colossal porkbarrel known as the Special Recovery
Capital Projects Program. Will the Minister confirm that the secretariat
administering that program does not operate out of his own Department
but, indeed, reports directly to Cabinet through Senator Keith Davey, and
that the key employees engaged in determining priorities under that pro­
gram are former government employees presently operating under con­
tract? If so, will he table for the examination of all Members an indication
of which of these employees have long distance telephone credit cards, car
rental cards, government codes, expense accounts, and the like? In other
words, will the Minister indicate the extent to which government
employees are being paid massive contract amounts to administer this pro­
gram if they are trusted political friends of the Government?
Hon. Donald J. Johnston {Minister of State for Economic and Regional
Development and Minister of State for Science and Technology): Mr.
Speaker, I will begin with reference to the porkbarrel. It is curious that the
Hon. Member for Richmond-South Delta would raise that since there is a
very significant project in his own riding at the Vancouver airport. I look
to the riding of the Hon. Member for Vancouver Centre. I see she is not
with us today. There is something like $41.9 million in that porkbarrel for
Vancouver Centre. Looking on from there, I see Vancouver Quadra, and
I think the Hon. Member for Vancouver Quadra might be here, $15 mill­
ion in Vancouver Quadra. I go over to the Island to Esquimalt-Saanich.
The Hon. Member for Esquimalt-Saanich is here. He is not blushing, I
hope, about receiving $8.5 million for the graving dock improvements in
Esquimalt. I see the Hon. Member for Hillsborough sitting here, to go to
the other end of the country, $14.5 million for the veterinary college in
Charlottetown. I do not hear him saying that he would not like the project
there. I note that the Leader of the Opposition is not present. It is interest­
ing that we have a $25.9 million project in Central Nova, Ferguson Indus­
tries Ltd. of Pictou. Just so this will be completely non-partisan, I see the
Hon. Member for Winnipeg-North Centre is here.
Mr. Speaker: Order.
One interesting aspect of this debate is that it shows us how the ad
hominem as a form of argument in dialogue is intimately linked to the ask­
ing of questions in the context of the dialogue. Mr. Siddon's question
requests Mr. Johnston to "indicate the extent to which government
employees are being paid massive contract amounts ... if they are trusted
political friends of the Government." This question is a complex or multiple
question. It contains a conditional and some conjunctions. It is also a
loaded question, in that some of the propositions contained within its parts
24 INFORMAL FALLACIES

as presuppositions are damaging to the position of Mr. Johnston's party. It

presupposes that "massive" amounts are being paid to "trusted political
friends of the Government." This presupposition is hardly one that the Lib­
eral Party could accept without appearing to be guilty of patronage or
"porkbarrelling," as the process of giving favors to your political friends is
called.
Note also that the question is constructed in an unfairly aggressive
way. It asks for specific amounts of these expenditures. However, if Mr.
Johnston were to provide these figures, thereby giving a direct answer to
the question posed, he would automatically be incriminating his own party.
He might well take issue with the highly prejudicial presumptions of the
question as posed. But if he provides a direct answer, he automatically
becomes committed to assent to those presuppositions.
Consequently, we need to ask here: in the context of the dialogue, is
Mr. Johnston obliged to directly answer the question? Or would it be fair if,
instead of giving a direct answer, he parried the question by asking another
question, for example, or by challenging some of its presuppositions? Given
the context and the specifics of the question, it seems fair and reasonable
that some of these latter options should be open without Mr. Johnston's
reply being open to reasonable criticisms of unfairly evading the question or
giving an irrelevant answer (ignoratio elenchi fallacy). When Mr. Johnston
then proceeds to reply to the question by means of an ad hominem rebuttal
to it, we should judge the legitimacy of his refutation of the attack posed by
the question in the light of the latitude that should be judged permissible
for a fair reply to the question.
Mr. Johnston's reply is very clever. Instead of giving the contract
amounts for the project queried, he systematically goes around to several
opposition members, including Mr. Siddon, and gives the amounts each of
them has been granted for significant projects in each of their own home
ridings. He says of one of these members, "I do not hear him saying that he
would not like the project there." What is Mr. Johnston suggesting by this
clever reply? He is saying that if the special capital recovery project could
be called a "porkbarrel" then so, equally well, could each of these other
opposition projects be called "porkbarrelling." It is a classical tu quoque
reply of ad hominem argument. Instead of answering the question directly,
Mr. Johnston turns the question back upon its asker and his supporters.
Is this ad hominem rebuttal fallacious? This judgement can only be
resolved, I suggest, in light of the question it retorts to. Is the question
 A NEW MODEL OF ARGUMENT 25

unfairly or aggressively loaded? If so, the ad hominem reply need not be

automatically dismissed as evasive or diversionary.
Indeed, it could be fairly argued that the question itself poses an ad
hominem attack against Mr. Johnston's party. To answer the question
directly is to concede "massive contract amounts" paid to "political
friends." Hence the question itself poses a form of personal or circumstan­
tial attack on the conduct of the party in power. In this light, Mr.
Johnston's reply can be seen as an ad hominem reply to an ad hominem
attack. It is truly a tu quoque rejoinder.
It should be pointed out that Mr. Johnston did not give a direct answer
to the question. His reply could possibly be criticized — perhaps fairly to
some extent —■ on those grounds. Yet given the aggressive nature of the
question, it seems fair and reasonable that he should be allowed to begin his
answer by rebutting some of the prejudicial presuppositions contained in
the question. His ad hominem reply does precisely that, and very effec­
tively. In effect, Mr. Johnston's reply makes the point that payment of
funds to one's political supporters is not necessarily a "porkbarrel." If the
funds are fairly paid, through due processes, with no conflict of interest or
clearly culpable ulterior motive, then such contracts or projects need not be
labelled by the perjorative term "porkbarrel." In fact, it is a separate ques­
tion to be established whether such contracts are in some clearly established
way illicit, and may therefore be described as a "porkbarrel." Mr. Siddon's
question — so Mr. Johnston's reply alleges — had not given evidence of the
illicit nature of the contract in question. Therefore Mr. Siddon's question,
which describes the contract as a "porkbarrel" within the question itself —
really begs the question, so to speak, of whether it is truly a "porkbarrel"
practice. Mr. Johnston's reply then, in effect, rebuts that presumption by
pointing out that one might equally well call, tu quoque, any of the opposi­
tion's recently funded projects "porkbarrels." He is turning the presump­
tion of the question back on its poser. He is, in a nutshell, questioning the
question. To the extent that the question is unfairly loaded, we can say that
Mr. Johnston's reply to it is fair and reasonable, and therefore commits no
ad hominem fallacy.21
Notice what has happened here, however. We have two parties in the
debate. One poses an ad hominem attack and the other marshalls a clever
ad hominem reply. Either the questioner or the answerer could be criticized
for various reasons, in the context of the dialogue. Conventionally, ad
hominem has been thought of as a fallacy, an argument that seems valid but
26 INFORMAL FALLACIES

is not. Our evaluation of this dialogue challenges that whole conception of

argumentum ad hominem as a fallacy.
By the analysis given in chapter 9, ad hominem is a form of criticism in
dialogue that, as a move in argument, can be in some cases permissible and
reasonable. Ad hominem, then, can be a non-fallacious argument in some
cases. In this case, whether a fallacy is committed depends on how the ad
hominem rebuttal fares as a reasonable response, given the structure of the
question that elicited it. Argument must be seen in the context of such an
evaluation, not just as a set of propositions where one follows deductively
from the others, arbitrarily called "premisses." Argument must be seen as
a dynamic relationship where one arguer's position is challenged or
attacked, and the other arguer's move is seen as a response to this chal­
lenge.
Hence we are led again to the main thesis of this monograph that fal­
lacies in realistic arguments can only be fairly evaluated by viewing the
argument as part of a logical game of dialogue. At one level, we have realis­
tic arguments like the parliamentary debate above. But to evaluate the
moves in these realistic arguments as "fair" or "fallacious" criticisms, we
need to have a model of dialogue as an abstract normative structure against
which the real argument can be evaluated.

6. Basic Categories of Argument Study

On some occasions in a dispute about a certain contention, neither
party may have access to additional evidence that would be sufficient to
persuade the other party to come over to his side of the issue. In such a
case, one's own argument may not be able to be strengthened, by building
it up with positive, new evidence. An alternative in this type of stalemate is
to adduce arguments to weaken your adversary's case for his side of the
contention at issue. For if his arguments are weakened, then your own side
may prevail. When the arguments are nearly enough balanced, yet neither
side wins and there is no new evidence for the moment that would tilt the
scales one way or the other, an argument to refute or undermine your
opponent's argument may be the only option.
If the above type of situation reflects the reality of argumentation, a
distinction needs to be made between two kinds of argument. One kind of
argument is the argument based on evidence that purports to be universal,
on verified knowledge that anyone should accept. This kind of evidence is
sometimes called reproducible evidence, meaning that it can be verified by
 A NEW MODEL OF ARGUMENT 27

any observer in like circumstances. A second kind of argument is based on

premisses accepted by some particular party to the argument. This basis of
argument does not purport to be universal. It represents the position of a
specific arguer, which that arguer accepts. But others may not share this
viewpoint. An argument of this sort is always relative to what this particular
arguer should accept. This kind of argument is essentially dialectical — it is
an argument devised by one party and directed against (or in some cases
towards) the position of another party in an interpersonal dialogue.
The most basic kind of ad hominem fallacy consists in the confusion
between these two species of argument. It is an argument of this form: your
dialectical argument for proposition ρ is weak, or based on premisses that
can be shown to be inconsistent, therefore ρ must be false, or at least not
based on verified, universal knowledge. This kind of move from the dialec­
tical to the reproducible argument is a fallacious kind of move. Although an
arguer's dialectical case for ρ may be weak, it need not follow that there is
no reasonably strong reproducible evidence for p.
To clarify the basic distinction involved here, we need to define our
terms more precisely. We shall define a dialectical argument as essentially
two-person in just the following sense. Let Ρ be the premisses and  the
conclusion of an argument. An argument 'P, therefore C' is dialectical if,
and only if, it is a move or series of moves, in a game of dialogue, advanced
by one participant with the objective of proving his thesis. Usually Ρ will
comprise propositions that are commitments of one or the other player.
Usually  will be an interim conclusion of the player who advances the
argument, that is, a proposition leading in that player's strategy towards the
proof of his own thesis. An argument 'P, therefore C' is monolectical if, and
only if, the set Ρ purports to be universally acceptable, reproducible evi­
dence not restricted to propositions that are commitments of a particular
participant in the argument. Very often Ρ will be some scientific or empiri­
cal evidence based on experiment, or some other form of reproducible ver­
ification.
Our next few remarks on categories and classifications of arguments
apply to both monolectical and dialectical arguments. However, our con­
cern is mainly with the latter. Therefore, our examples and cases will be
drawn from the dialectical model of argument.
There are two basic functions of argument. One is proof, or what
might be called positive proving, and this refers to proving your conclusion,
based on premisses that establish that conclusion, or contribute to its estab-
28 INFORMAL FALLACIES

lishment. In dialectic, proof normally means proving your own thesis from
your opponent's premisses. The other function of argument is refutation,
where an argument falsifies or undermines a conclusion. In dialectic, refu­
tation normally means showing that your opponent's position is inconsis­
tent.
To say that an argument is proven because it has not been refuted
could be a dialectical form of the ad ignorantiam fallacy, though such an
inference is not always fallacious. Similarly, in monolectical terms, to say
that a conclusion is falsified because it has never been verified could be one
form of ad ignorantiam fallacy. Much more needs to be said about ad
ignorantiam arguments, but it is clear that the distinction between positive
proof and refutation lies at the heart of that informal fallacy.
In a game of dialogue where classical logic provides the rules of valid
argument, proof and refutation amount to nearly the same thing. If the
rules of valid argument include addition, simplification, and disjunctive syl­
logism, then any conclusion follows validly from an argument with inconsis­
tent premisses. Hence proving that one's opponent's commitment-set is
inconsistent, in a game of dialogue that includes these three logical rules,
amounts within the next few moves to proving one's own thesis from that
opponent's commitment-set. This well-known fact is essentially theorem 1
of Walton (1984, p. 140). However, in some non-classical propositional cal­
culi — like relatedness logics and relevance logics — theorem 1 does not
hold. Hence in games with logical rules based on these non-classical logics,
refutation and proof are farther apart.
Ignorance of refutation {ignoratio elenchi), in the Aristotelian tradi­
tion, refers to a failure to show that one proposition is contrary to, or
refutes another. This failing of argument refers to a badly mounted refuta­
tion which fails to demonstrate inconsistency. Failure to prove the conclu­
sion an arguer is supposed to prove, in a game of dialogue, is a parallel fal­
lacy.
The traditional informal fallacies pertain to different types of argu­
ments and criticisms that have gone amiss in one way or another. Fallacies
may be related to three basic kinds of failures of an argument.
1. Wrong Premisses. In the characteristic case, in games of dialogue, this
failing means that the premisses of one party's argument are not all
commitments of the other party. In other cases however, this criticism
may involve an allegation that the party criticized has too many (re­
dundant) premisses, has too few premisses, or has the wrong sort of
 A NEW MODEL OF ARGUMENT 29

premisses for other reasons, e.g. premisses that are not topically
related to the conclusion.
2. Invalid Argument. This criticism is familiar in the usual approaches to
logic. It means that the conclusion does not follow from the premisses
by the logical rules. It represents a failure to prove.
3. Wrong Conclusion. The problem here is that the argument could be
valid, as required by 2., yet have the wrong conclusion. In games of
dialogue, it is required at the outset that all participants have a specific
proposition designated as their thesis (conclusion) to be proven. In
realistic argumentation, this requirement may not be met, or not
clearly met.
The above three types of failures of an argument are basic to all studies of
fallacies and applied logic. They are somehow fundamental. They can be
applied to monolectical arguments as well as dialectical.
However, there are various other failings of argument that can play a
role in reasonable criticisms of arguments. These include the following.
4. Questions Badly Asked. Questions have presuppositions and therefore
the act of questioning itself may be subject to reasonable criticisms.
Questions can be complex. Questions can be unfairly aggressive and
prejudicial to the answerer's position. Therefore, the fair management
of question-answer relationships is an important part of procedures
that must be regulated in games of dialogue.
5. Failure to Answer Questions. Since a question may be unfair, or
unreasonably posed, it should not always be required in rational
dialogue that an answer must provide a direct answer to every ques­
tion. Yet the answerer should be constrained to give a reasonable
answer without being unduly evasive. Failure to answer a question may
be criticized, in some circumstances as committing a fallacy of irrele­
vance.
6. Failure to Argue. Sometimes an argument fails, not by having the
wrong sort of premisses, but by not having any premisses at all. The
basis of some criticisms is that the arguer has created an emotional
"smoke screen" in lieu of providing any propositions that can clearly
be identified as premisses.
7. Too Many Arguments. Another form of failure of argument is where
more than one argument is advanced. Equivocation is a case in point.
Because of an ambiguous term in one or more of its propositions, an
30 INFORMAL FALLACIES

"argument" may really be several arguments rolled up into one schema

or bundle that looks like a single argument. The problem is that the
different arguments may exhibit different failings, and thereby confuse
the one to whom the argument is directed. Another interesting failure
of this sort occurs where one arguer advances an argument meant to be
taken as inductively correct, whereupon his critic criticizes it as deduc­
tively invalid. There may be different standards or types of argument,
and these may be mixed up.
There are other failings of argument that are connected to the criticisms
that form the basis of the traditional lore of informal fallacies. But these
seven fundamental kinds of argument failure provide a structural
background against which a good deal of the criticisms at work in the fal­
lacies can be studied.
It is now time to look at the fallacies and other relevant aspects of
argument in greater detail. We begin with the first group of fallacies intro­
duced, namely the fallacies that have to do with "hot" appeals to emotion.
Before getting into the detailed workings of the fallacies, we conclude this
chapter by making a brief general statement of our objectives to be carried
out in the ten chapters to follow.
The subject-matter of informal fallacies is a pragmatic testing ground
for formal theories of argument and dialogue. The three main formal
theories of logical dialogue — games advanced in recent times are those of
(1) Lorenzen and his school (the Erlangen school), 22
(2) Hintikka 23 and his students, including Carlson,24 and
(3) Hamblin 25 and his followers, including Mackenzie. 26
Also, the theory of argument advanced by Barth and Krabbe 27 has devel­
oped the formal structure of the Lorenzen school. The problem for the
study of the fallacies is to apply these three theories to the various special
contexts of argument required to make sense of the informal fallacies in
attempting to fairly and reasonably judge — and provide general standards
to judge — criticisms, refutations and arguments of a realistic sort that rep­
resent traditional "fallacies" and other significant phenomena of the world
of fallacies and argumentation. All three of these theories can be fruitfully
applied to the traditionally chaotic and undisciplined domain of the fal­
lacies. But in certain special contexts, one theoretical approach may offer
clear advantages over the remaining pair. Each theory has its advantages.
And, given the undeveloped state of the study of the fallacies, it would be
premature to declare that one of these theories is the best for all or most
 A NEW MODEL OF ARGUMENT 31

purposes. That is not our objective, at any rate.

My objective is to look over enough of the rich domain of the fallacies
to allow a working, pragmatic model of argument to emerge. Having
already sketched out the very rough edges of the model in this first chapter,
the subsequent aims are to justify the applicability of that model in relation
to the fallacies, and to explore its refinement in building particular models
of questioning, criticism, rebuttal, and so forth, for various special contexts
of reasonable argumentation interchange.

NOTES

1. See Johnson and Blair (1980).

2. See Rescher (1976) for a systematic study of arguments where shift in the burden of proof
is significant.
3. See Hamblin (1970).
4. The most complete survey of the fallacies is that of Hamblin (1970, chapter 1). A less com­
prehensive introduction is given in Walton (1984, chapter 1).
5. Works of Hamblin (1970) and Walton (1984) bring out this lesson in depth.
6. See Walton (1980a).
7. So we will argue in chapter 9, at any rate.
8. See Woods and Walton (1982, chapter 5).
9. Chapter ten of this monograph is an extended study of the fallacy of equivocation.
10. See, for example, Zadeh (1975) for structures appropriate to this sort of reasoning.
11. See Woods and Walton (1977).
12. See Woods and Walton (1978).
13. An extended study of this example is given in Walton and Batten (1984).
14. A more extended discussion of a case similar to this one is to be found throughout Walton
(1984).
15. See Walton (1982).
16. See Woods and Walton (1977).
17. I would like to thank Tan Bee Chin for drawing this debate to my attention.
18. Subsequent extended discussion of ad hominem criticism in chapter 9 will clarify just
when an ad hominem attack becomes fallacious.
19. Hamblin (1970) emphasizes this aspect.
20. I would like to thank Marc Frechette for drawing this debate to my attention.
32 INFORMAL FALLACIES

21. However, the reader who has doubts about this evaluation might wish to go on to chapter
9 to get a fuller appreciation of the ad hominem argument as sometimes being a reasonable
move.
22. Lorenzen (1969).
23. Hintikka(1979).
24. Carlson (1983).
25. Hamblin (1970) and (1971).
26. Mackenzie (1981).
27. Barth and Krabbe (1982).
 CHAPTER TWO: HOT RHETORIC AND ARGUMENT

The first group of fallacies we discussed in the previous chapter were

the "hot" appeals to emotion. These fallacies are worth discussing as a
group because conversational quarrels and debates are heavily interlaced
with emotional appeals. Some have even suggested that political debates,
for example, are largely decided on the basis of emotional factors. We have
already seen how the ad hominem attack is a powerful and central refuta­
tion in political debate. In this chapter, we will concentrate mainly on appe­
als to force, pity, and popular sentiment as strategies of emotional persua­
sion, though we include ad hominem argument for study as well. However,
our indepth study of ad hominem argument will not be carried out until
chapter 9.
Emotion is intimately involved with the objective of rhetoric. For
rhetoric is the attempt to persuade a specific, target audience. And an emo­
tional relationship is a personal link between individuals. To become
involved in an emotional relationship, two people concentrate their atten­
tion on each other, they "build a wall around themselves and exclude
others" (Bailey, 1983, p.48). An emotional relationship suggests that we
suspend calculation and logic, that we trust and grant credit to each other.
According to Perelman (1982, p.17), persuasive discourse is addressed to a
person's unthinking reactions, as opposed to convincing discourse which
appeals to reason.
A longstanding problem with the fallacies that trade on emotion is that
they have always seemed to be quite separate from any standard structures
used by formal logic. However, we will see that our new models of argu­
ment and dialogue structures show promise of assisting us in the analysis of
these heretofore elusive fallacies. Let us therefore turn to a closer study of
some of these famous emotive fallacies.

1. Appeals to Popular Sentiment

The ad populum fallacy is usually characterized, we noted in the last
chapter, as an argument that appeals to mass enthusiasms or popular senti-
34 INFORMAL FALLACIES

ments in order to win assent for a conclusion. However, the initial observa­
tion in relation to this characterization is that there is nothing intrinsically
wrong in appealing to mass enthusiasms in an argument. Indeed, in a demo­
cratic society, such mass appeal will be positively required if one's argu­
ment is to be successful in winning over an electoral majority. Thus the
logic textbooks usually add that the ad populum type of argument becomes
fallacious when the emotional appeal is used as a diversion to disguise the
argument's failure to be backed up by proper evidence. By these lights
then, the ad populum is really a fallacy of irrelevance.
The most common type of example cited by the textbooks as ad
populum reasoning is that of commercial advertisements. For example,
commercial messages for life insurance tell us very little about the com­
plexities of the different kinds of policies that are available. Instead the
commercial commonly appeals to general attitudes of an average family
using phrases like "peace of mind today" and "security for the future." The
commercial message, for example, may be placed against scenes of a family
happily engaging in everyday recreational activities on a river bank. These
activities include fishing, and praising the insurance company as a place
where the family and the insurance agent can "work things out together."
The friendly insurance agent also happens to be fishing on the same river
bank. Clearly such a scene is concocted specifically in order to appeal to
popular sentiment. The insurance agent is portrayed as being a typical,
even lovable sort of person, who goes fishing and has the same aspirations
as most of us.
Of course there need be nothing wrong per se, as we have already
noted, in such an appeal to popular sentiment. However, what one suspects
might be questionable about the commercial is its exclusive preoccupation
with the emotional impact of the appeal, and the conspicuous absence of
any attempt to tell us about the relative merits of the insurance policies that
the company makes available, the capabilities of the company, or any of the
more "relevant" aspects of insurance policies that should properly play a
role in our decision to buy insurance from this particular company. How­
ever, it can't be denied that the folksy appeal of the commercial message
does indeed succeed in winning the acceptance of its target viewership and
in helping to induce them to buy insurance from this company.
A similar example that could be cited is a commercial advertisement
for a lumber company which stresses a theme message of "back-to-the-
land" and "do-things-for-yourself." The character in the commercial is an
 HOT RHETORIC AND ARGUMENT 35

appealing handyman do-it-yourself person who pleasantly conveys feelings

of accomplishment and pride of building things for yourself. By creating an
emotional appeal to the popular sentiment of personal pride in building
something for yourself, this commercial advertisement certainly conveys a
popular message that does probably do quite an effective job of winning
over customers. But one's suspicion is aroused by the observation that facts
about the quality of the tools or lumber sold there, which might enable one
to intelligently arrive at a decision about whether to consider buying those
tools or lumber, is entirely missing from the message. Thus the provision of
adequate evidence that would enable a rational consumer to decide
whether or not to buy these products has simply been evaded in place of a
popular appeal to personal accomplishments.
More subtle types of ad populum appeal can occur in advertisements
for automobiles or electrical appliances, where, for example, there may be
legitimate facts about technical specifications and other truly relevant infor­
mation mixed in with the appeal to feelings. Including this genuine informa­
tion sometimes has the psychological effect of heightening the persuasive­
ness of the appeal. For example Goebbels pointed out that audiences found
falsified and heavily emotional patriotic appeals of reports from the battlef-
ront were believable if some general factual information was mixed in with
the lies and propaganda. 1
One can certainly appreciate the widespread ethical deficiencies of
advertising practices that are reflected in mass emotional appeals. There­
fore it is easy to concede the importance of the ad populum as a fallacy that
does have wide influence in persuasion. But we have to be extremely care­
ful when characterizing this type of persuasion as a fallacy. For a fallacy, as
we have noted, is a fallacious argument. Therefore we have to be very clear
on precisely what is fallacious about an emotional appeal to popular senti­
ments. What is fallacious then about such an appeal? The main answer
given by the textbooks as we noted above, is that the popular appeal may
be a diversion from correct argument. That is, we seem to be confronted
with a fallacy of irrelevance. So the main question for the ad populum is
this: precisely what is meant by irrelevance in this context? However, there
is a previous point that should be considered as well. Since many an ad
populum appeal seems to comprise mainly a visual and emotional appeal to
sentiment, can we truly say it is an argument? For example, does the
demagogue even argue at all when he whips his audience into mass
approval by theatrical means of playing on the audience's emotions? Our
36 INFORMAL FALLACIES

comments on the use of popular appeal in advertisements suggest that the

fallacy may consist rather in the avoidance of argument altogether.
In the lumber company example the point has been evaded by appeal
to the pride of personal accomplishment. Perhaps then something evasive
or manipulative has transpired, but where is the fallacy and where is the
argument? What seems to be wrong is that argument of any sort has been
foregone in favor of a direct appeal to sympathies and attitudes. We have to
ask: where precisely are the premisses and conclusions in such an appeal? If
they are not there to be found, then there is no argument, therefore there
can be no fallacy. Yet perhaps it could be that the fallacy consists in the
evasion itself. What could be wrong is the lack of argument through the
irrelevance of the emotional appeal, a kind of deception by distraction.
Many textbooks support this view that what is fallacious about the ad
populum is that the emotional appeal is irrelevant to the argument. But
does this mean that the emotional appeal is an irrelevant argument, or does
it mean that it is no argument at all, and therefore nothing relevant to the
argument has been given? So the point remains that if, in fact, there is no
identifiable argument that is given — no set of propositions that makes up
the argument — then there cannot be a fallacy or fallacious argument,
except by default, by reason of the lack of argument. Our lumber company
has ostensibly committed a fallacy by changing the subject from the quality
of their products to the popular folklore of back-to-the-land fantasy. But is
a fantasy a statement or an argument? If not, it remains questionable
whether the ad populum appeal can genuinely be analyzed as a fallacious
argument. A fantasy may be effectively used by an advertiser who wishes to
glamorize his product by sketching out a daydream for us that excites our
approval and admiration. That fantasy may make a statement of sorts but
can we literally assume that such a fantasy can be broken down into a
specific set of statements with determinant truth values? In the examples
given above it is by no means obvious how such an analysis could be carried
out. To that extent then we should remain reluctant to acquiesce too readily
in the assumption that the ad populum appeal is a fallacious argument. 2
There are three kinds of cases to be analyzed. If the emotional appeal
to popular pieties cannot be stated in propositional form, as a set of premis­
ses, then there may be no argument at all, only the semblance of argument.
In that case, we need to ask: can the avoidance of argument be fallacious?
Much depends here on how we undertake to define 'fallacy'. Yet in the con­
text of a game of dialogue, the failure to engage in the game by making an
 HOT RHETORIC AND ARGUMENT 37

appropriate response could certainly qualify as a good reason for a fair

judgement that a participant should lose the game. Hence evasion of argu­
ment, or refusal to argue, could be accounted a kind of fallacious step in
dialogue — a failure to follow the agreed-upon procedure. This is the sixth
fault of argument listed in chapter 1 — the failure to argue.
The second kind of case is that where the emotional appeal can be
stated as a proposition or set of propositions. Here we have an argument in
the sense of a set of premisses and a conclusion. Whereas the fault in the
first type of case is that there are no real premisses at all, the problem here
could be that there is simply too little in the way of good premisses to estab­
lish the arguer's conclusion. It could be a case of the first type of fault of
argument, that of wrong premisses. Instead of advancing reasonable pre­
misses to prove his conclusion, the arguer goes on with many weak premis­
ses that appeal to popular solidarity as a basis for his argument.
The third kind of case is one where the arguer may have some reason­
ably good premisses in his argument, but may interweave popular emo­
tional appeals with those premisses. Here the problem is to try to sort out
the factual premisses from the emotional baggage. As our examples of
advertisements using ad populum strategies showed, the most effective ad
populum arguments tend to be of this sort. The failure here is the seventh
fault of argument — too many arguments.
Other kinds of faults may be associated with ad populum arguments as
well. The appeal to emotion may be, in some cases, a straying off the topic
set as the issue of the argument.
One worrisome factor is that underlying these various faults, the basic
thing to be remembered is that an appeal to popular sentiments may not be
fallacious. Although, as we have seen, the ad populum appeal is associated
with several faults of argument, there is nothing intrinsically wrong, in
some cases, in appealing to popular sentiments. Hence care is needed to
not leap too quickly to conclude that any popular appeal is an ad populum
fallacy.
We have previously claimed that the legitimate objective in a logical
game of dialogue is to prove your conclusion from the other arguer's pre­
misses. Therefore, you must and should argue from premisses that your
target audience does or will accept. In some cases, given this objective of
argument, it is not only legitimate but positively appropriate to use premis­
ses that are highly acceptable, i.e., popular pieties, to your target audience.
Hence we must stress that, in some cases, there may be nothing unreasona-
38 INFORMAL FALLACIES

ble about an argument that may, on the surface, appear to be ad populum.

In many cases however, the worst fault of the ad populum appeal is the
attempt to avoid argument altogether by identifying both sides of the
dialogue with the position of some mutual interest group. It's as if the
arguer says: "We are both true believers in a common cause. The others,
whom we must exclude, are the common enemy or evil parties. Let's both
belong to this group and not worry about 'truth' or 'falsehood'. Nothing
matters but our common group solidarity."
Bailey (1982, p.134) warns of the kind of rhetoric that compartmen­
talizes the moral world into true believers and outsiders, insisting that truth
is all on the one side. As an example (p.134) he quotes a speech of Walter
Reuther from 1957 on the subject of racketeers in trade unions.
I think we can all agree that the overwhelming majority of the leadership
of the American movement is composed of decent, honest, dedicated
people who have made a great contribution involving great personal sac­
rifice, helping to build a decent American labor movement .... We happen
to believe that leadership in the American movement is a sacred trust. We
happen to believe that this is no place for people who want to use the
labour movement to make a fast buck ...
As Bailey points out, the intent of this speech is to put its message beyond
all doubt and questioning by offering it as a fact to which everyone must
agree. The listener is left no choice. There is no room for argument left.
Hence the whole purpose of argument as reasonable dialogue is undercut at
the outset. There is meant to be no need for real dialogue in this type of ad
populum rhetoric. The "we" who believe in the "movement" are included.
The "people" who want to make a "fast buck" are excluded. That is the
whole message. Either join in the labor movement position, or be excluded
as an infidel — that is the only choice offered to the listener. Hence Bailey
calls this type of ad populum argument a "rhetoric of belonging" (p.135).
Clearly it is not a form of reasonable dialogue, but an attempt to avoid the
necessity for argument as reasonable dialogue. It lacks the give-and-take of
interpersonal dialectic.
We conclude that although appeal to popular sentiment need not
necessarily be, in itself, fallacious, it is often associated with one or more
faults of argument in a characteristic way. Anyone attempting to evaluate
arguments ought to be aware that appeal to popular sentiment is very often
a signal for one or more of the faults. In some instances then, a case can be
made out for judging ad populum argumentation to be fallacious.
 HOT RHETORIC AND ARGUMENT 39

2. Appeals to Force
The ad baculum, as we saw, is the fallacy traditionally said to be com­
mitted when one appeals to force or the threat of force to make one's audi­
ence accept a conclusion. A typical sort of example cited by the textbooks
would be the following: a large heavy-set man comes to the door, knocks on
the door and says to Dagwood, who opens the door, "I am selling this win­
dow cleaner." Poking Dagwood in the chest, he says, "And I'm not a guy
who likes to fool around. Either you buy it or I'll punch your lights out!" In
the next frame we see Dagwood walking back into the house with two bot­
tles of window cleaner, saying, "He has a very persuasive sales approach."
Here it is not too difficult to appreciate the force of the ad baculum in per­
suasion. However, one has to be very careful to ask whether such an appeal
is clearly an argument. Consider the salesman's utterance, "Either you buy
it or I'll punch your lights out!" Here is a disjunction, the first disjunct
being an imperative and the second being, to all appearances, a statement
of intention. Certainly there is a difference between an imperative utter­
ance and a statement. Statements are true or false, whereas imperatives do
not clearly or at least directly take truth values as properties. Rather, they
are satisfied by the action or inaction of the person to whom they are
addressed. Is the salesman then really constructing an argument or is his
speech-act of some other type? The use of the connector word "or" suggests
argument, but on the other hand, the use of the imperative expression
"You buy it!" suggests that if there is argument, it is certainly not a "fac­
tual" argument of the more usual sort.
Another point we should keep in mind is that appeal to force may not
always in itself, be logically incorrect. If two Prime Ministers are arguing
over whose country is more powerful, and one is able to show that his coun­
try has a greater number of tanks, aircraft, etc., the statement of this Prime
Minister may in fact constitute an appeal to force, yet it may have quite a
legitimate role to play in the diplomatic negotiations. Thus, although what
we might have here is in some sense an appeal to force, there is a good
question about whether or how such an appeal to force is fallacious.
Another example is the law that gives as a penalty for conviction of
drunk driving the suspension of one's driving license. Such a law may con­
stitute an appeal to force or threat to force, or perhaps even to fear, but
nonetheless one would hardly want to dispute the legitimacy of such a law.
Certainly it is not clear why such a law would constitute a fallacious appeal
to force. 3 We conclude that an appeal to force need not be in itself falla-
40 INFORMAL FALLACIES

cious. One must be very careful in approaching such cases to attempt to

precisely determine what is thought to be fallacious.
Certain kinds of reasoning in natural language have to do with practi­
cal decision-making about actions. In the context of practical reasoning,
certain kinds of inferences may be reasonable even if they do not conform
in a self-evident way to the standards of classical deductive logic.
Examples of practical reasoning can be drawn from the literature on
the theory of action.
One view of practical inference is that the agent has some goal or
intention and contemplates undertaking some action necessary to bring
about that goal. Foremost among the exponents of this approach is von
Wright who proposes the following schema for practical inference (1972,
p.48). E is an end, A an action, and X an agent.
X intends to make it true that E.
X considers that unless he does A, he will not achieve this.
Therefore, X sets himself to do A.
The sort of example von Wright (1963) has in mind is the following. X
intends to meet his fiancee in London. Unless he takes the next train, he
will not meet her. Therefore, X will take the next train. This sort of exam­
ple of reasoning is quite plausible as an instance of an inference we com­
monly make. But there are many puzzling aspects of it. If the conclusion is
an action, as von Wright's schema suggests, then the inference is not deduc­
tively valid. For even if X truly intends to meet his fiancee, and taking the
train is necessary to achieve this, there might be any number of reasons why
X might in fact fail to take the train. Accordingly, von Wright considers
various modifications of this basic schema.
Whatever precise form the ultimate expression of this schema might
take, let us call it the necessary condition schema of practical inference. By
contrast, other philosophers have contrasted on what we might call the suf­
ficient condition schema. Anscombe (1957) was one of the first in recent
times to draw our attention to the existence of this form of argument in sev­
eral interesting passages of Aristotle.
The type of example Anscombe and Aristotle thought most important
is one of this sort: this medication is effective to cure condition C; this
patient has condition C; therefore this medication is indicated for this
patient. One wonders here whether the conclusion of this practical infer­
ence should be an action or some form of policy or recommendation to
 HOT RHETORIC AND ARGUMENT 41

carry out a certain type of action. As with the necessary condition schema,
it remains unclear with the sufficient condition schema what sort of infer­
ence is involved. At any rate, parallel to von Wright's initial schema, the
following tentative form of the sufficient condition schema seems appropri­
ate.
X intends to make it true that E.
X considers that if he does A, he will achieve this.
Therefore, X sets himself to do A.
Again, whether the conclusion should state that X will do A or that X sets
himself to do A or that X should do A remains unclear.
Despite these unclarities and uncertainties about the precise form of
the practical inference, a fair presumption in this type of inference may
sometimes be "reasonable" or "correct," even if it is not clear whether or
when it is deductively valid. At least, it would be unwarranted to claim that
all instances of practical reasoning are "fallacious" or "incorrect."
Now the problem posed for us here is that if first-person practical infer­
ences can be reasonable in some sense, then third-person practical infer­
ences that take the form of a warning — advice to someone to avoid
adverse consequences — could also be a reasonable type of inference. For
example, I might say to you: "If you persist in carrying out course of action
A, disaster will ensue. Therefore, you should avoid persisting in carrying
out course of action A." On the model of practical reasoning, such a warn­
ing might be reasonable, and not in itself "fallacious" in a given instance.
Now the ad baculum threat is supposed, on the traditional account, to
be a fallacy. But here is the problem: what is the difference between a
threat and a warning? If a warning can be a reasonable form of advice,
according to the standards of practical reasoning in the context of decision­
making about actions, when does a reasonable warning become a "falla­
cious" threat in argument?
I don't claim to have a solution to this problem. I simply want to point
out that it should be regarded as a very serious problem by anyone who
would undertake to give an analysis of ad baculum as a fallacy. The solution
to this problem should be sought in the theory of speech acts in argumenta­
tive discussions developed by van Eemeren and Grootendorst (1983). What
is the difference between the speech act in which a threat is issued and the
speech act that comprises a warning to avoid certain consequences. Com­
mon intuitions tell us that the distinction is there to be made, but it is not
42 INFORMAL FALLACIES

obvious what guidelines should be appealed to.

This problem is a realistic one for the dialectic of ad baculum argumen­
tation because anyone who issues a threat by the ad baculum argument will,
of course, deny that he has made a threat when his ploy is challenged. In
realistic argumentation, the easy way out is to ingenuously claim that a
reasonable warning was only intended. But if we do not know how to tell
whether a particular speech act was really a threat or merely a warning,
how could a criticism of ad baculum fallaciousness even be nailed down?
The charge could never be made to stick if this escape route is always avail­
able in reasonable procedures of dialogue.

3. Appeals to Pity
A third type of argument in this class of irrelevant emotional appeals is
the fallacy of ad misericordiam or appeal to pity. The textbooks commonly
cite cases where the defendant in a criminal trial brings his wife and chil­
dren into the courtroom, and the lawyer takes up a strategy of playing
directly on the emotions of the jury, instead of trying to put up good argu­
ments for the innocence of the defendant. The suggestion here is that the
emotional distraction is really beside the point, and therefore commits a fal­
lacious ad misericordiam appeal.
However, a study of actual examples of appeals to pity reveals that it is
often not too clear whether, or to what extent, such appeals are fallacious.
An advertisement for the Foster Parents Plan of Canada shows a picture of
an appealing but pathetically distraught young boy around five years of age,
underscored by a large headline, OSMAN KNOWS HUNGER AND ALL
ITS PAIN. This little boy is described as hungry, hopeless and confused.
The ad says that he knows hunger and pain but does not realize that there
is any other way of life. Then the appeal goes on, "By becoming a foster
parent you can help a child overseas. Your support will provide food, clo­
thing, shelter, and so many more children wait. Please — complete the
coupon below, or call us toll-free." The picture and the headline are very
definitely a direct appeal to pity, but is it fallacious?
Hamblin (1970) joins us in pointing out that in many contexts a propo­
sition is presented primarily as a guide to action. So the same sort of prob­
lem of practical reasoning we found in the ad baculum is present here as
well. Where action is concerned it is not so clear that pity is irrelevant. Pity
is a Christian virtue, and in this case could perhaps be as good a reason as
one might like for contributing money to the Foster Parents Plan of
 HOT RHETORIC AND ARGUMENT 43

Canada.
What then could be fallacious if anything, about such an appeal? The
traditional move of the textbooks with this alleged fallacy, as with others we
have seen, is to claim that the fallacy occurs when the emotional appeal is
used as a diversion. The fallacy, it is urged, is one of irrelevance. The
suggestion in this particular case then may be that the appeal to pity, while
reasonable in itself, falls short of correct argument insofar as it fails to sup­
ply details of precisely how the funds solicited will be spent.
With any appeal from a charity it is, of course, quite reasonable to ask
what percent of the funds are actually getting to the alleged recipients as
against the percentage of funds that are being used for administrative or
other purposes. Any potential donor who is experienced with the workings
of charitable organizations should request or expect some assurances about
the efficiency of the organization in achieving its desired objectives. If,
however, there is an entire absence of such information and instead merely
a heart-tugging picture is printed and some remarks directly appealing to
pity, then there is a good question whether the issue is being evaded.
In this particular case, however, it seems hard to say. Should the Fos­
ter Parents Plan of Canada advertisement have included more information,
or at least some information about how the money is administered, and pre­
cisely what happens to your contribution? Or were they justified in exclu­
sively directing their advertisement towards a direct appeal to the charitable
instincts of the reader? The answer depends on what the conclusion of the
argument really is. Therefore perhaps the lesson at this point is that one
must be very careful in analyzing the alleged fallaciousness of such appeals
to first of all determine precisely what are the statements that supposedly
make up the premisses and the conclusion of the argument.
Consider as another curious example the following argument. Premiss:
the seal hunt is a bloody, brutal murder of baby animals. Conclusion: there­
fore [to protect the species from extinction] the hunt must come to a stop.
What is interesting in this particular argument is that if the bracketed part
is omitted, the argument may not be fallacious. However, notice that if the
bracketed part is included then the objection is much more plausible that
the premiss is an irrelevant ground for arguing to the conclusion. In such a
case then one must be extremely careful to attempt to discern initially what
is the issue; what propositions are being disputed; what is the conclusion
that the arguer is allegedly putting forward as his or her thesis. Having
determined this, then we can work towards an analysis which may enable us
44 INFORMAL FALLACIES

to arrive at some estimate of the extent to which the premiss is relevant.

However, in the absence of a clear statement of the conclusion, any judge­
ment of the fallaciousness of such an appeal based on relevance founders
for lack of information.
The problem is that in many natural language disputations, particularly
in ordinary quarrels or debates, there is no clear agreement on the part of
the participants precisely what the issue is to be disputed, or what is the
proposition that each participant is set to prove as his or her conclusion. In
the absence of this prior determination, it is not possible to adjudicate on
appeals to pity, fear or popular sentiments.
In relation to the question of whether such an appeal may or may not
be incorrect or fallacious, Kielkopf (1980) reminds us that too often stu­
dents are encouraged to acquiesce in the superficialities of the standard
treatment by alleging that a fallacy has occurred as soon an any appeal is
made to the emotions of pity or popular piety, whether fallacious or not.
The tendency widely encouraged by the textbooks is to cite a fallacy of ad
misericordiam or ad populum as soon as it becomes evident that there is any
emotional appeal being made in an argument. However, as we have seen,
such a move may be presumptuous and unfair given that emotional appeals
are at any rate sometimes reasonable enough.
Another problem with these emotional fallacies is posed by the fact
that each has now been characterized according to the standard accounts of
the textbooks as being at bottom, a case of irrelevance or diversion. On the
other hand, each constitutes, it appears to be claimed, a distinctive type of
irrelevance. The ad populum is defined as irrelevance by popular appeal.
The ad baculum is defined as irrelevance by appeal to force. The ad mis­
ericordiam is defined as irrelevance specified by appeal to pity. Therefore
we have to ask this question. Are these fallacies all instances of one fallacy,
namely the fallacy of irrelevance — often traditionally called the ignoratio
elenchi fallacy — or is there some clear basis upon which we can differen­
tiate amongst these three as being distinct fallacies in their own right? Now
the problem here, taking the ad misericordiam as an example, is that the
emotion of mass enthusiasms or popular sentiments is not a factor in argu­
ment that we would know how to identify with clarity and precision in any
kind of logical model or by means of precise guidelines. In fact the emotion
of appeal to mass enthusiasms seems to be a psychological factor or quan­
tity. Now we have to be quite clear here that the question we are asking is
whether the ad misericordiam is a fallacy in the sense of it being an incor-
 HOT RHETORIC AND ARGUMENT 45

rect argument. That is, we want to ask precisely when is such an appeal
incorrect or wrong, normatively as a type of clearly identifiable argument,
without at least directly raising or involving the question of when such a
type of argument is effective rhetorically or persuasive psychologically. The
problem is that there could be some danger in trying to identify the ad mis-
ericordiam as a specific type of fallacy by identifying it by means of the pre­
sence of a psychological factor of mass enthusiasms. Presumably our task is
the logical, or at least the normative task, of trying to identify precisely
when such a type of argument is incorrect or invalid. Can we say that such
an argument is incorrect, although it may be irrelevant, precisely because it
commits the specific type of irrelevance identified with mass enthusiasms?
Are we thereby asking whether there can be a specific set of criteria
that will enable us to determine when an argument appeals to mass
enthusiasms? This question seems somewhat misguided. It is really an
empirical matter or perhaps a psychological matter when an argument is an
appeal to mass enthusiasms as opposed to, say, an appeal to the emotion of
fear or the threat of force. If this is correct, it may be then that all three of
these types of emotional appeals are really sub-classes of one specific type
of fallacy, namely that type of argument that goes wrong by virtue of irrele­
vance, and that a way of differentiating between these three types of irrele­
vance involves matters more of a pragmatic than a purely logical character.
Thus the question here is — are there really three fallacies or just one fal­
lacy with three different psychological manifestations? This is a good ques­
tion and it tests what we mean by the expression "informal fallacy."

4. Overly Personal Argumentation

Another interpretation of what is fallacious about the ad populum
argument is that such an argument is directed to a specific audience rather
than being an argument from objective premisses. It is sometimes pointed
out that the ad populum arguer adopts a strategy of selecting premisses
specifically so that they will be accepted enthusiastically by the audience
that is being addressed. The fallacy here would be scheming to convince an
audience by appealing to assumptions that appear tolerable to that audi­
ence rather than arguing from premisses that are known to be true, or at
least that can be shown to be true independently of their appeal to a par­
ticular audience. Perhaps what appears wrong with such a strategy is its
subjective orientation which subverts the goal of arriving at the truth by
choosing premisses independently of their appeal to the particular audi-
46 INFORMAL FALLACIES

ence.
In analyzing this interpretation of the ad populum however, one must
be very careful not to accept a questionable presumption that the only
legitimate function of argument is to reason from premisses that are known
to be true. As we have already pointed out, sometimes legitimate argu­
ments in dialogue proceed from premisses that are conceded or accepted,
but that are nevertheless not certainly known to be true.
A properly dialectical perspective on the concept of argument should
allow arguments from premisses that may not be known to be true. We
need to be reminded that there need be nothing illogical per se in arguing
from presumptions that may be plausible to a participant in the argument
but yet may not be known to be true or even true with high probability.
However, we have already stressed the dialectical model of argument as
our central concern in this monograph.
Perhaps what this interpretation of the ad populum argument alludes
to is, however, a different kind of fallacy which Johnson and Blair (1977,
p.158) called the fallacy of popularity. This sort of argument occurs in two
forms.
F l . Everyone believes p, therefore ρ is true.
F2. No one believes p, therefore ρ is false.
As Johnson and Blair point out, F l . and F2. are forms of arguments that
stated as they are, would not be likely to fool most audiences. However,
they could be forms of argument that are implicit in fallacies that are closely
related to the ad populum. Certainly it could be highly fallacious to start
with a dialectical premiss like 'Everybody believes that p' or This audience
believes that ρ quite enthusiastically' and proceed to conclude directly that
ρ must be true for every possible audience. However, this fallacious move
in itself would not seem to represent, at least exclusively, what is meant by
the ad populum fallacy because in the forms F l . and F2., there need be no
emotional appeal to popular sentiments or mass enthusiasms in a particular
case.
However, this interpretation of the ad populum fallacy does raise the
interesting question of to what extent it is legitimate to reason from premis­
ses concerned with a specific arguer or group of participants in argument.
The question is: what is the relationship between dialectic and truth?
The same sort of problem is posed by a fourth kind of fallacy that also
often pertains to emotional diversions. This is the ad hominem argument or
 HOT RHETORIC AND ARGUMENT 47

"argument against the man." Modern texts, as we saw, standardly distin­

guish between two forms of ad hominem, the circumstantial and the abusive
variety. The abusive ad hominem was said to be committed when an arguer
attacks his opponent directly instead of trying to disprove the opponent's
argument. This form of argument is often called "character assassination."
As with the previous three fallacies, however, one must be very careful
here. Sometimes in a court of law, for example, it might be quite reasona­
ble to question the witness's character or the defendant's personal charac­
teristics or memberships in groups or parties, evidence that might be pre­
judicial against the case. On the other hand, often such an appeal to per­
sonal presumptions about the arguer's character can be clearly specious and
unfair. As with the previous three fallacies, in order to attempt to sort out
between the correct and the incorrect instances, the notion of relevance
may be appealed to, the claim being that the argument on the basis of the
opponent's personal characteristics is incorrect when there is an evasion of
correct argument or genuine evidence, and instead personal vilification is
substituted as an emotional appeal. In this respect then, the ad hominem is
quite similar to the previous three emotional fallacies, and insofar as it is
claimed to be a fallacy of irrelevance, similar problems are posed in its
analysis.
The circumstantial ad hominem consists in the attempt to refute a per­
son's argument by alleging an inconsistency between the opponent's thesis
and some action brought about by the opponent or other circumstances
pertaining to the person of the opponent. Most typically a circumstantial ad
hominem allegation consists in the accusation that the arguer does not prac­
tice what he or she preaches. For example, a driving instructor tells you not
to drink and drive because it leads to accidents. Upon inquiry you discover
that he himself recently had an automobile accident as a result of drunken
driving. You might well argue that his argument is worthless because he
himself is guilty of the very kind of act he counsels against.
The classic example of the circumstantial ad hominem is an argument
we will call the sportsman s rejoinder.4 When accused by a critic of sacrific­
ing innocent hares or trout for his amusement, the sportsman replies to his
critic, "Why do you feed on the flesh of harmless cattle?" Commonly this
reply is said to commit a fallacy because the sportsman does not try to
prove it is right to sacrifice animals for his amusement, but rather dwells on
the critic's own circumstantial inconsistency posed by the fact that the critic
himself eats meat. Thus according to Copi (1972, p.76), this fallacy is really
48 INFORMAL FALLACIES

a failure of relevance. "Arguments such as these are not really to the point;
they do not present good grounds for the truth of their conclusions but are
intended only to win assent to the conclusion from one's opponent because
of the opponent's special circumstances." Copi seems to be suggesting that
an argument presents good grounds for the truth of its conclusion and is
therefore a correct argument only if its premisses are really to the point,
that is, relevant. However, precisely what relevance consists in, in such cor­
rect arguments is something we are not told by Copi, nor do other standard
textbooks on fallacies tell us very much about precisely what type of rele­
vance could be meant here.
The circumstantial ad hominem is clearly an interesting, yet at the
same time, difficult conception to analyze. It may well be true in argument
that dwelling on inconsistency in an opponent's position is unfair if it is an
evasion from evaluating propositions that may directly bear on the oppo­
nent's thesis as conclusion. Yet on the other hand it may seem in some
cases quite reasonable that if the opponent has committed himself to a posi­
tion that is internally inconsistent one should attack the inconsistency. On
the other hand, typically the type of inconsistency cited in an ad hominem
allegation is not a logical inconsistency but rather a pragmatic sort of incon­
sistency between the propositions that the arguer has asserted and the
actions or other background circumstances of that arguer. The fact that
other than direct logical inconsistencies are involved also leads to certain
complications in pinning down and adjudicating ad hominem allegations.
We return to the important ad hominem fallacy in a later chapter to con­
sider it in much greater detail.
All the fallacies we have looked at so far in this chapter involve the
lively interpersonal exchange characteristic of the personalistic and emo­
tional processes characteristic of quarrels and debates. Of course, in the
ordinary conversational quarrel, whether domestic or public, the lack of
clear guidelines or rules amounts to the anarchy of argument. At that level,
the fallacies remain as elusive as ever. However, we know that organized
debates are often regulated. Could the debate provide a medium of argu­
ment appropriate to the study of fallacies?

5. The Rhetorical Debate

The principle objective of a participant in forensic debate is to win by
a majority vote or by judgement of the referee. The referee or the majority
characteristically arrive at a ruling by means of rules for debate which will
 HOT RHETORIC AND ARGUMENT 49

be stipulated in advance and made known to the participants. Thus the

debate is regulated, and thereby has an important advantage over many
quarrels and disputes in natural argumentation that may be entirely unregu­
lated. In many an ordinary quarrel, as we all know too well, there are not
specific rules that permit an adjudication on who wins or loses specific
moves in the argument. Nor are there rules constituting what constitutes a
correct or incorrect argument, nor rules concerning clear violations of pro­
cedural mechanisms for ensuring fair disputation. However, the debate
often does have clear rules to regulate these functions and thereby initially,
at any rate, seems to have some quite decisive advantages over the ordinary
quarrel. By means of these rules the debate would appear to have an ele­
ment of objectivity that is conspicuously lacking in many an ordinary quar­
rel.
Because it is partially regulated, and yet partially unregulated, a
debate can therefore very often quite effectively bring out the arguments
for and against a position. Thus the debate can serve the course of proof
and correct argument in an effective way.
However, it should always be kept in mind that debate is adversarial in
the sense that its primary object is for the participant to win even at the
expense of correct argument. Thus in debate, the argument that serves
effectively to convince the audience or referee, and thus contribute to the
winning of the debate, may be fallacious from a logical point of view and
yet still rhetorically effective as an argument. Therefore an argument that is
successful in a debate — successful in the sense of attaining its rhetorical
objective of persuasion — may be logically incorrect and fallacious in itself.
This possibility suggests that debating techniques as instruments of reason­
ing should be viewed with considerable caution. In fact, there are sound
reasons for believing that there is considerable divergence between rhetori­
cally effective debating technique and correct logical reasoning.
In debate the contending debaters put the proposition at issue to test
by questioning and answering. Ideally all the participants in the debate can
ask probing questions, and answerers must try to provide effective answers.
However, what constitutes a rhetorically effective answer may not necessar­
ily be a relevant answer or a logically correct answer. Nonetheless, as the
debate progresses certain alternatives that may have been considered as
policies or proposals will drop from contention and no longer be supported.
These proposals will fail the test of scrutiny by participants in the debate.
Perhaps in debate, if it is successful, a consensus will be reached. Very
50 INFORMAL FALLACIES

often this consensus is a compromise which has arisen out of the opposed
contentions of the debate. The proposition that survives is then one thought
likely to be worthy of political support and action. But is that survival due
to its logic?
We must ask whether the process of debate might still tolerate quite a
degree of deviance from logical correctness in the arguments that are used
to arrive at that conclusion. A traditional defence of the debate as an instru­
ment of responsible reasoning is through what we could call the survival of
the truth hypothesis. This hypothesis is that truth is more highly valued than
falsehood in the free marketplace of disputation, and that therefore truth
more often survives the destructive forces of opposition and criticism than
falsehood. Defenders of the cogency of debate have thereby hypothesized
that truth has a natural tendency to prevail in most cases, if the debate has
been sufficiently thorough and fair. This is an optimistic point of view on
the rhetorical debate as an instrument of logical reasoning.
The practical problem with the hypothesis is that rarely are parliamen­
tary debates or legislatures free markets of opinion and argument. Par­
liamentary debate is too often plagued by the inattentiveness of the partici­
pants, and too often the issue receives insufficient attention during the heat
of a contested issue to really give the argument its due careful analysis so
that fallacies and other inexactnesses are detected or avoided. Such a fail­
ure is not simply an indication of the laziness of parliamentarians, but may
be due to practical constraints of time and party and caucus discipline.5
Members of parliament are rarely free to be rationally persuaded to
accept or reject a position. Rather they are required to adhere to a party
line and to reject arguments that might give the opposition an advantage,
even if that rejection involves sharp practice or overlooking of some finer
points of logical precision. Very often nowadays, parliamentary debates are
televised and therefore there is considerable pressure upon the debaters to
cut a good figure in persuading the broad mass of viewers to certain courses
of action which are consistent with one's official party policies. In such a
rhetorical setting the participants are often driven to ploys and strategems
which, while scarcely logical, may effectively serve the purposes of the
exercise provided only that they persuade the target audience or defeat the
opposition. Thus a fundamental shortcoming of the debate is that its audi­
ence-directed adversary structure causes it to be essentially open to a sys­
tematic acquiescence in a dangerous form of ad populum reasoning. For the
very function and purpose of the debate is to win over the broad majority
 HOT RHETORIC AND ARGUMENT 51

to a certain course of action. Thus the objective of the parliamentary

debate has the model of ad populum reasoning built into it as a condition of
its very success and rhetorical effectiveness in argumentation.
From the logical point of view this ad populum feature is a most impre­
ssive shortcoming of the debate. Legal debate in criminal and civil trials is
of course legally regulated, and in this regard has similar favorable charac­
teristics to the argumentation of parliamentary debate. Nevertheless legal
argumentation, like parliamentary debate, has as its goal the defeat of the
opposition. The structure of legal argumentation is again essentially a
rhetorical and adversary procedure. The objective the attorney has in mind
then is the effective, legal defeat of his opponent's arguments regardless of
the intrinsic correctness of his own arguments or in the incorrectness of the
opponent's arguments. The goal is not to be "logical" as such. If there is a
conflict between logical reasoning and persuasion of the jury then persua­
sion of the jury must be given priority, if one is to play the game and win.
Therefore, exactly as in the political debate, the model of ad populum
reasoning is also built into the success conditions of legal argumentation.
Ultimately the goal must be to rhetorically convince the jury and if this goal
involves the emotional use of the ad hominem or ad misericordiam fallacy,
then such strategems are not only permissible but positively required in
order to effect the favorable outcome — that is, to win the case for one's
client. This sink-or-swim aspect is not just an incidental part of legal
argumentation but is inherent in the very structure of the process.
Thus it is clear that there is no guarantee that a sequence of legal
reasoning in adversary courtroom procedures that is rhetorically effective
as argumentation will thereby be logically correct and free from fallacies.
As with parliamentary debate, however, this is not to say that legal
reasoning is not very often illuminating, instructive, and consistent with its
own stated goals. For sometimes the best way to be perceived to be reason­
able is actually to be reasonable.
In chapter six we will look at a case study of legal disputation. Let us
now turn to an example of parliamentary debate.

6. Case Study: Parliamentary Debate

The following specimen dialogue from the House of Commons
Debates in Canada is quite interesting because it involves an attempted
refutation or reply to a speaker's question that combines elements of three
traditional fallacies, the ad hominem, ad verecundiam (appeal to authority),
52 INFORMAL FALLACIES

and the ignoratio elenchi (fallacy of irrelevance). Mr. Rae, the New Democ­
ratic Party shadow minister for finance, asked Mr. MacEachen, the Liberal
Finance Minister, whether reduced loan rates ought to be made available to
farmers and small businessmen, given that bank profit figures show recent
substantial increases. This question occurs in the second segment, under the
heading Banks and Banking, the transcript of an oral question period in the
debate. Mr. MacEachen replies to the question by referring back to a previ­
ous part of the debate where Mr. Rae had asked other questions. The trans­
cript of this part is the first part printed below, under the heading The
Economy. Mr. MacEachen's answer to Mr. Rae's question in part 2 is that
Mr. Rae had earlier contradicted himself in part 1, by first relying on Statis­
tics Canada as a source of information, and then saying that we should dis­
regard Statistics Canada as an unreliable source. The topic of Mr. Rae's
earlier arguments which were claimed to be inconsistent by Mr.
MacEachen's reply had to do with employment trends in the construction
industry. Hence there is a good question here just how relevant Mr.
MacEachen's reply is. Hence the possible involvement of the ignoratio elen­
chi fallacy, which has to do with irrelevant refutations.
But the ad verecundiam fallacy — the subject of chapter six — appears
to be involved as well, for Mr. MacEachen has accused Mr. Rae of a misuse
of the appeal to authority. Indeed, Mr. MacEachen has accused Mr. Rae of
an inconsistent appeal to the authority of Statistics Canada. Therefore,
since an allegation of inconsistent position is being maintained, there may
be some elements of an allegation of the circumstantial ad hominem fallacy
as well, a fallacy that has to do with refutation of an inconsistent position.
Let us look at the actual wording of the debate and then proceed to an
analysis of it.

1. THE ECONOMY
MEASURES TO MAINTAIN EMPLOYMENT
Mr. Bob Rae: Madam Speaker, my question is directed to the Deputy
Prime Minister and Minister of Finance. The figures released today by
Statistics Canada show that unemployment hits hardest those who are, in a
sense, in the front line of the government's interest rate policies, such as
construction workers, where employment is down by 6,000, and agricul­
ture, where employment is down by 2,000, to name just two industries.
Will the minister not now repudiate the view of Governor Bouey of
the Bank of Canada that labour markets are tight?
Oral Questions
 HOT RHETORIC AND ARGUMENT 53

Will he not now take measures which will change the policy before
interest rates start eating away further at employment?
Hon. Allan J. MacEachen: Madam Speaker, I draw to the attention of
the hon. member that although seasonally adjusted unemployment is up
slightly from last month, which was the best employment record in five
years, there is another figure which he or his party draws to our attention
occasionally, namely, the actual count of unemployment. The number of
unemployed decreased from April to May by 32,000.
These are important facts the hon. member ought to bear in mind,
plus the fact that in the last year the number of jobs created in Canada has
been the best of any industrialized country in the world. It seems to me
that these facts do not encourage anyone to take the action which the hon.
member recommends.
Mr. McGrath'. Tell that to the million unemployed.

STATISTICS OF UNEMPLOYED
Mr. Bob Rae: Madam Speaker, in the "Allan in Wonderland" world
in which the minister inevitably lives, he has managed not only to readjust
the employment rate seasonally, but in the case of the construction indus­
try, he has managed to adjust the seasons. We are now heading into the
winter downturn of the construction industry, which is quite an achieve­
ment for the minister.
The minister asked us to look at the real figures. One spokesman for
Statistics Canada was quoted this week as saying that if an interviewee
were asked the question whether or not he or she had worked during the
week, and that person had worked for one hour during the week, the per­
son would be classified as employed for the purposes of the statistics.
Does the minister not think it is time that the Statistics Canada survey
indicates what are the real levels of employment, under-employment and
enforced idleness in the economy? We know that in addition to the over
800,000 people who are listed in the figures, there are another approxi­
mate 500,000 who have been either laid off or lost jobs and are discour­
aged from seeking further work. If the minister wants to talk about the real
figures, does he not think a more realistic figure, or one that could be pro­
vided in the Statistics Canada survey itself, would be 1.35 million
unemployed, rather than the 845,000 unemployed figure which the govern­
ment released today?
Hon. Allan J. MacEachen: Madam Speaker, I do not agree at all with
the private statistical service which the NDP usually provides to us, when
economic indicators are rather strong, in order to distort the economic pic­
ture in Canada.
54 INFORMAL FALLACIES

Mr. Rae: Madam Speaker, if reliance on Statistics Canada is a private

survey, then it is something in which the minister himself engages more
than once a day.

2. BANKS AND BANKING

INCREASE IN PROFITS OF CHARTERED BANKS —

INCREASED CHARGES TO FARMERS AND SMALL
BUSINESSMEN
Mr. Bob Rae: In light of the fact that profit figures for the chartered
banks today indicated a substantial increase, up 49 per cent — an increase
which follows a 27 per cent rise for all chartered banks for the period
between 1979-80 — and since tax rates for the chartered banks are still at
20 per cent, will the minister now reconsider his government's rejection of
a policy which would require the chartered banks to set aside a fixed por­
tion of their financial portfolios for mortgage loans, loans to farmers and
small businessmen, at rates which bear some relation to their costs, and the
borrowers' ability to pay, so that we can stop this policy of driving people
to the wall and forcing idleness and real hardship on the Canadian people?
Hon. Allan J. MacEachen: Madam Speaker, I may have misun­
derstood the member's first question. I thought he drew upon Statistics
Canada and its information to demonstrate certain employment trends in
the construction industry. Then in his next question he said that we should
disregard Statistics Canada —
Mr. Rae: That is not what I said.
Mr. MacEachen: — upon which he relied in his earlier question,
because it is not reliable.
Mr. Rae: That is just misleading.
Mr. MacEachen: Maybe the hon. member would clarify his own
thinking before he puts questions to me on the orders of the day.
Looking over the whole sequence of argumentation, we can break Mr.
Rae's question-asking moves into three basic parts. In his first segment of
dialogue, Mr. Rae is asking about the topic of unemployment. He asks Mr.
MacEachen: (1) whether he will now repudiate the view that labour mar­
kets are tight, and (2) will he not take measures to change fiscal policy
before interest rates cause further unemployment. In response, Mr.
MacEachen challenges Mr. Rae's employment figures. In his second seg­
ment of dialogue, Mr. Rae goes on to challenge the methods used by Statis­
tics Canada to collect and tabulate official government employment figures.
 HOT RHETORIC AND ARGUMENT 55

Thus it was Mr. MacEachen who really brought up the question of the relia­
bility of employment statistics, and Mr. Rae's second question represents a
counter-attack. Finally, Mr. Rae's third question, as we already noted, asks
whether reduced loan rates ought to be made available to farmers and small
businessmen, given the increases in bank profit figures. This is a third dis­
tinct topic of question-asking, though not entirely unrelated to the subject-
matters of the two preceding questions.
Now we turn to the question of the fairness of Mr. MacEachen's reply
to Mr. Rae's third basic segment of question-asking dialogue. Two distinct
questions about this reply need to be distinguished.
(1) Is Mr. MacEachen's allegation of inconsistency basically accurate in
itself? That is, did Mr. Rae actually contradict himself at some points?
(2) Should it be acceptable for Mr. MacEachen, in his answer to Mr. Rae's
third question, to accuse Mr. Rae of contradicting himself whether in fact
his accusation was accurate or not — in his previous two questions? Was
Mr. MacEachen being in some sense "irrelevant" by failing to answer the
third question and instead launching an interesting, and perhaps evasive,
attack against parts of the previous dialogue? Neither of the questions can
be easily resolved.
The first observation that should be made about (1) is that Mr.
MacEachen's accusation is based on a correct proposition. There certainly
is a contradiction involved — Mr. Rae did draw on Statistics Canada infor­
mation, and then cast doubt on the information provided by Statistics
Canada. Mr. Rae's first two questions justify the attribution of both a prop­
osition A, and its negation 6 1A, as part of the position he adopts. In fact, it
is plausible to suggest that it is this underlying correctness of Mr.
MacEachen's allegation that makes it rhetorically quite effective as a refu­
tation of Mr. Rae's argument. Mr. Rae is made to look silly, and his reply
"That is just misleading," while perhaps accurate, is ineffectual. Clearly,
Mr. MacEachen has scored heavily in the debate.
But what of the logic of the dialogue? Should the refutation be treated
as a correct response to Mr. Rae's questioning? Classical logic is by itself no
help. In classical logic, an inconsistent set of propositions implies any prop­
osition you like. 7 In other words, classical logic provides very little direction
concerning how to handle inconsistency beyond ruling that an inconsistent
set of propositions cannot collectively be true. Should we simply say then
that Mr. Rae is refuted?
Before acquiescing too quickly, we should check to see if both A and
56 INFORMAL FALLACIES

IA are propositions about the same subject-matters. For it could be that

Mr. Rae is justified in relying on Statistics Canada for information on one
topic and condemning Statistics Canada's inaccurate information on
another unrelated topic. Statistics Canada, for example, might have reliable
banking statistics on interest rates, yet quite unreliable statistics on employ­
ment trends. So, at any rate, might one consistently claim.
The general lesson here is that simple, topic-neutral inconsistency of A
and 1A is not enough to refute one's position, where that position may be
on different subjects. The question of relevance may intrude.
However in this case Mr. Rae is squarely caught. His reliance on
Statistics Canada in his first question was specifically on the topic of
employment figures. Whereas his questioning of the reliability of the
methods used by Statistics Canada to collect information was also on the
subject of employment figures. Hence the stated inconsistency is directly on
the same topic. Hence Mr. Rae is refuted fairly enough as far as question
(1) is concerned.
To turn to (2) now, was Mr. MacEachen justifiable in adducing the
inconsistency of the first two questions as a response to the third question?
This question is more difficult to rule firmly on. As far as the conventions
of the rules of debate for the House of Commons Debates are concerned,
Mr. MacEachen's reply would appear to be acceptable. The Oral Question
Period allows quite a latitude in the answering and posing of questions, and
if Mr. MacEachen chooses to recast his remarks back to previous questions,
that is his privilege. However, concerning the fairness of the disputation
apart from specific parliamentary conventions, is there still room for criti­
cism?
Some would point out that in fact Mr. MacEachen never did answer
Mr. Rae's third question. Was his final reply therefore not a successful eva-
tion of the question, and open to criticism on that count? The observation
is a good one, but the topic of the third question is not entirely unrelated to
those of the first two, having also to do with economic statistics. Moreover,
if Mr. Rae is inconsistent in his evaluation of statistical sources, does not
Mr. MacEachen's refutation weigh heavily against Mr. Rae's whole argu­
ment, and not just the first two parts exclusively? So there are two sides to
this issue.
That the issue is non-trivial for the logic of dialogue is suggested by the
general remark that very often some will argue that ad hominem refutation
of a position by allegation of internal inconsistency of the opponent's cir-
 HOT RHETORIC AND ARGUMENT 57

cumstantial position is essentially an irrelevant move. Reason: the real issue

of whether the opponent's argument, or some consistent part of it, offers
genuine evidence for his conclusion is not thereby addressed or resolved.
For some, ad hominem refutation is always, in a certain sense, beside the
point of the argument. For others, ad hominem argument is legitimate and
relevant because it does evidence a weakness in the arguer's overall posi­
tion on which his argument is based. This is not the place to resolve this
issue, but merely to note that it is a genuinely interesting general point con­
cerning the logic of ad hominem refutations.
To sum up then, Mr. MacEachen's refutation is correct as far as (1) is
concerned. But although it is partially correct in regard to (2), there could
be also some grounds for questioning it as a relevant answer in relation to
the concerns of (2).
We are now in a position to evaluate the analysis of this debate given
by the Informal Logic Workbook (1981, p.73).
In answering Mr. Rae's question, Mr. MacEachen does not address
the issue of interest rates. Instead, Mr. MacEachen initiates a shift in the
focus of debate. He chastises Mr. Rae for using Statistics Canada informa­
tion when Mr. Rae had previously criticized the accuracy of Statistics
Canada's unemployment rate. Mr. Rae, however, did not indict the credi­
bility of Statistics Canada as an institution, but only challenged the
unemployment rate, the accuracy of which he feels is debatable. The thing
to be noted here is that Mr. Rae does not cite Statistics Canada as the
source of his information concerning bank profits. In addition, Mr. Rae
only questioned the accuracy of the unemployment rates, an issue which
was not even remotely related to the profits of some of the chartered
banks. Mr. Rae's questions are therefore not inconsistent.
Debating Mr. Rae's use of Statistics Canada information is irrelevant
and does not answer Mr. Rae's question concerning government policy
and interest rates. It constitutes a Red Herring because it shifts the focus
of the exchange, and takes place in an adversary context (the House of
Commons debates). By changing the subject, Mr. MacEachen instigates a
shift of focus in the exchange, and fails to defend his government's position
against Mr. Rae's implied criticism that it should be changed. (False
Charge of Inconsistency, Red Herring).
The thrust of this allegation of irrelevance (Red Herring) given in the sec­
ond paragraph is, as we have seen in our remarks on (2), partially justifi­
able. However, the specific allegation of irrelevance made in the top para­
graph is based on a mistaken interpretation of Mr. MacEachen's argument.
According to the Workbook account, Mr. Rae does not cite Statistics
58 INFORMAL FALLACIES

Canada as a source of information concerning bank profits, and he only

questioned Statistics Canada on the accuracy of employment rates. Because
bank profits and employment rates are, according to the Workbook, "not
even remotely related," Mr. Rae's questions are not inconsistent. As we
noted above, there could be a legitimate question about the relatedness of
subject-matters in evaluating this dialogue. However, as we saw above, Mr.
MacEachen does not allege that Mr. Rae's third question-asking segment
about banking is inconsistent with his second question-asking segment
where he berates the accuracy of Statistics Canada's methods of collecting
employment figures.
As the examination of the whole dialogue clearly shows, Mr.
MacEachen is citing an inconsistency between the first two question-asking
discourses of Mr. Rae. I think it is quite clear what Mr. MacEachen refers
to when he speaks in his rejoinder under 2. of Mr. Rae's "first question"
and his "next question" respectively. He refers to Mr. Rae's first dialogue
under 1. as the "first question," and then Mr. Rae's long three-paragraph
dialogue starting with the "Allan in Wonderland" reference as his "next
question." In both these questions, as we showed in our analysis above, the
topic is the same, namely having to do with employment trends and statis­
tics.
The problem here is that the Workbook analysis has failed to get Mr.
MacEachen's arguments right by concentrating too exclusively on the
dialogue under 2., and thereby failing to correctly account for the preceding
context of dialogue here fully quoted under 1. The lesson is that any critic
must be very careful to first of all be clear on what the argument is. What
are the questions and propositions that actually make up the dialogue? The
way many realistic dialogues and debates are, it may be very hard to tell.
And therefore this initial step of analysis is non-trivial. Here we see instruc­
tively how a lapse in getting the argument well-located can result in a griev­
ous skew of correct analysis.
So Mr. Rae is indeed inconsistent in a way that is really damaging to
his argument. His lapse of good argument was adroitly picked up by Mr.
MacEachen, and provided a basis for him to avoid directly answering Mr.
Rae's third question. The refutation works because the alleged inconsis­
tency is on the same topic — employment figures — as the overall dialogue,
noting especially the first part, makes clear. This does not mean that Mr.
Rae's argument is worthless altogether or completely unrepairable. Given
an opportunity to respond fully, perhaps he could explain why his reliance
 HOT RHETORIC AND ARGUMENT 59

on certain Statistics Canada figures is not undercut by his questioning of

certain aspects of their methods. Too often however, the pressure of time
and party discipline in parliamentary debate does not allow for fully
adequate reflection and response.
Another factor that should be mentioned in connection with (2) above
concerns the nature of Mr, Rae's third question. If a question is unfairly
"loaded," e.g. "Have you stopped beating your spouse?" it may be accept­
able for an answerer not to directly answer it.8 Instead, the answerer may
call for the question to be clarified, or he may question or challenge a pre­
supposition of the question. In short, sometimes failure to directly answer a
question is fair practice in dialogue.
Mr. Rae's question is a loaded one, at least in the sense that a negative
answer implies that the answerer will not reject "a policy of driving people
to the wall and forcing idleness and real hardship on the Canadian people."
Surely the loaded nature of this question could offer justification for Mr.
MacEachen's avoidance of a direct answer.
The way Mr. Rae's question is phrased implies that the government's
rejection of the policy Mr. Rae proposes has driven people to the wall and
forced hardship on the Canadian people. To pack such a conjecture into the
question presumes far too much and makes the question unfairly aggres­
sive. The language of "driving people to the wall" is also unduly emotional
(ad misericordiam).
Another interesting question about this dialogue is this: Mr.
MacEachen's answer refuted Mr. Rae's first two questions by showing an
inconsistency, so to what extent is it reasonable for an onlooker to assume
that the answer refutes (or answers) the third question as well? Now note
that in Mr. Rae's third question, no mention is made of the source of his
figures, nor is it indicated or not whether the claims made about bank pro­
fits are based on information from Statistics Canada.
Thus there is possible danger of fallacy here for the uncritical reader or
listener who may assume that Mr. MacEachen's reply is such a telling blow
that it challenges or even refutes the third question directly. Such an
assumption is one we must be careful to avoid.
Another critical point is that Mr. MacEachen's first answer (in part 1.)
does not match up perfectly with all of Mr. Rae's first question. Mr. Rae's
figures pertained to unemployment in construction and agriculture,
whereas Mr. MacEachen's figures given in his answer are general statistics
for all occupations. So there need be no conflict, for it is perfectly consis-
60 INFORMAL FALLACIES

tent for unemployment to be heavy in one sector, but not heavy overall dur­
ing a certain period. Nonetheless, Mr. MacEachen's response is basically
correct in that it is addressed to Mr. Rae's last question of his first dialogue,
pertaining to the effect of interest rates on unemployment (generally, we
may presume).

7. Conclusion
We can now easily see why the parliamentary debate is a fascinating
data bank of specimens of the fallacies so far identified and other interest­
ing moves in argument. But we have also seen the grave difficulties in trying
to pin down justified allegations of fallacious argument in such debates.
As we look at specimen examples of parliamentary debate and legal
debate in real life and in this book, many questionable moves that are made
in the heat of the adversary struggle can be revealed later — with all the
advantages of hindsight — in an objective perusal of logical analysis. It is
easy to denigrate these political debates and say that the participants are
simply stupid or inept. However, we must recall that it is one thing to
analyze and dissect a debate at one's leisure, using logical tools of analysis
to comb through the printed sentences, whereas it is quite another thing in
the heat of debate to struggle and contest with opposing parties in an unruly
adversary procedure, given also the previously noted constraints and limit­
ing objectives of parliamentary and legal debates.
We conclude then that these constraints and limitations mean that
there will always tend to be a good deal of slippage or skew between the
logically correct argument and the effective political or legal adversary
debate. It is not that we should be disappointed by the astounding illogical­
ity of debates when we look at them in hindsight. It is rather that the intrin­
sic objectives built into the very nature of the debate itself must mean that
the proceeding of the debate along precisely logical lines is not something
to be expected or hoped for. The coincidence of successful argumentation
in debate with purely logical lines of reasoning is at best a partial overlap­
ping.
Many very ordinary quarrels and disputes are hopelessly subjective
and therefore while they are often quite interesting and revealing as sources
of material on the fallacies, the problem is that there are no objective
guidelines that will allow us to determine in a fair and unbiased way
whether an argument is definitively correct or fallacious. The problem is
that if participants in argument do not formulate in advance clear and well
 HOT RHETORIC AND ARGUMENT 61

defined rules concerning what constitutes a good argument and what consti­
tutes the procedural rules for conducting the argument, then it will not be
possible to pin down fairly and precisely when one participant in argument
has committed a fallacy against another.
In the parliamentary debate and the legal adversary argument how­
ever, there is more hope for some objective regulation of the argument just
because law courts and parliamentary debates are partially regulated. The
problem, however, as we have seen, is that such regulation is of a weak
nature. Although the asking and the answering of questions is regulated,
nevertheless the regulation is not clear and precise enough to effectively
rule out fallacies nor is the type of regulation involved precisely designed
for that end. On the contrary, although it is designed to allow a favorable
interchange of ideas and conflict of viewpoints, nevertheless it quite openly
allows the fallacious argument to win conclusively.
The debate then is open to the fallacies we have examined because of
its essentially subjective nature. Where can we turn for an objective
account of what constitutes a correct or incorrect argument? Let's first look
to formal logic, to see what it offers. Another question that looms large —
what could we possibly mean by 'relevance?' The next chapter addresses
both questions.

NOTES

1. Jacques Ellul, Propaganda, New York, Knopf, 1972.

2. For a fuller elaboration of these points, see Douglas N. Walton, 'Why is the Ad Populum
a Fallacy?' Philosophy and Rhetoric, 13, 1980, 264-278.
3. For further discussion, see Kielkopf (1980).
4. A very interesting account of the sportsman's rejoinder was given by Richard Whately in
his Elements of Logic (New York, William Jackson, 1836). As we will see, Whately's account
was criticized by DeMorgan (1847). We return to a fuller treatment of the sportsman's rejoinder
in chapter seven, part two.
5. Elaborations on some of these shortcomings of the debate as a model of argument are
given in Woods and Walton (1982, chapter 2).
6. Negation is defined more fully in the next chapter.
7. Chapter two explains why.
8. This point is returned to in greater detail in chapter three, part three.
 CHAPTER 3: THE LOGIC OF PROPOSITIONS

In the last chapter we pursued the conception of argument as a lively

and often fractious process of personal exhortations and emotional inter­
relationships between arguers or between a speaker and audience. This vie­
wpoint is heavily laden with interpersonal psychological and sociological
dynamics. Perhaps our problem is that it is too rich and unmanageable as a
place for logic to get much of a grip on argument.
The viewpoint of formal logic, a mathematical discipline, takes a very
austere conception of argument as its central preoccupation. From this vie­
wpoint, an argument is a set of propositions, entities that have the proper­
ties of being true or false, never mind all that unruly interpersonal dynamics
of claim and counterclaim.
A big issue for us will be the applicability of such an austere, formalis-
tic approach to the realities of actual argumentation in natural language.
But before tackling such heady questions, it is well to review the basic ele­
ments of formal logic.
Deductive formal logic has to do with arguments like this one. If Bob
moved his pawn, then he is in check. Bob moved his pawn. Therefore, Bob
is in check. This argument is said to be deductively correct in the sense that
if both premisses really are true, then the conclusion must also be true.
There is no weaseling out of it. The argument's correctness is a matter of its
logical form because replacement of its two component propositions, 'Bob
moved his pawn' and 'Bob is in check,' with any other pair of propositions,
still retains the correctness of the argument. For example, the following
argument, because it has the same form, must also be deductively correct.
If Socrates is a man, Socrates is mortal. Socrates is a man. Therefore, Soc­
rates is mortal. Let us turn to an outline of the basic elements of classical
deductive formal logic.

1. Deductive Validity
What is an argument? Some would think our theory of argument in
chapter 1 much too generous, and would incline to a narrower conception.
64 INFORMAL FALLACIES

From the point of view of classical deductive logic, an argument is a set of

propositions. The propositions in an argument are divided into two sub­
sets, the premisses and the conclusion. An argument is said to be deduc­
tively valid (correct) if it is logically impossible for the premisses to be true
while the conclusion is false. Deductive logic is a kind of fail-safe procedure
— it is designed in such a way that it can never take you by a valid form of
argument from true premisses to a false conclusion. If you start with true
propositions, then you must only go to true propositions by valid deduc­
tion.
Consider the following argument. This parrot is either dead or asleep.
If it is dead, it will not be bought by Studs Terkel. It is not asleep. There­
fore, it will not be bought by Studs Terkel. Logic, in its classical guise, does
not tell you whether the premisses are true or not. In fact, it is quite possi­
ble that the parrot could be faking unconsciousness, and thus be neither
dead nor asleep. And logic does not tell you that the premiss is true, that
the parrot is not asleep, for example. Probably we may base our knowledge
of this premiss on some empirical finding, e.g. we prodded the parrot and
it did not respond as if it were awake. However, logic certainly assures us
that if the argument is valid and the premisses are true, then the conclusion
has to be true too. The argument is deductively valid in just this sense: it is
not logically possible for both premisses to be true and the conclusion false.
As far as classical deductive logic goes, it is quite enough for validity of
an argument that the truth of the premisses be sufficient for the truth of the
conclusion. Thus it is no disparagement of the validity of an argument if one
or more of the premisses happens to be false. Consider this argument.
Either Dick Cavett is a professional football player or a professional hockey
player. If he is a professional football player, he has a wooden leg. If he is
a professional hockey player he has a wooden leg. Therefore, Dick Cavett
has a wooden leg. Now notice in this argument that all the premisses are
false. And as we all know, the conclusion is false. Nonetheless the argu­
ment is certainly valid in the sense we just defined. For if the premisses
were true, the conclusion would certainly have to be true too. Such an argu­
ment could never take us from true premisses to a false conclusion.
The lesson here is just because an argument is deductively valid, that
does not mean it is as good as it could be in all possible respects. As we
have seen, it might have false premisses, and therefore in light of its premis-
sary worth, it could be a bad argument. It could also be viciously circular.
The argument, 'Dick's hair is brown, therefore Dick's hair is brown' is
 THE LOGIC OF PROPOSITIONS 65

deductively valid. Certainly this type of argument could never take you
from true premisses to a false conclusion — it doesn't even go anywhere.
But it is not an argument that is as good as one could like in all possible
respects. In fact, if you asked me to prove that Dick's hair is brown and I
advanced the argument, you would justifiably think it a very bad argument
indeed. Yet it is deductively valid.
Indeed, the above way of defining validity has the curious consequence
that any argument with inconsistent premisses must be valid. If the premis­
ses contain an inconsistency like 'Dick is in Berkeley' and 'Dick is not in
Berkeley/ then the premisses can never be true. So you could not possibly
go from true premisses to a false conclusion. Hence by default, as it were,
such an argument must be deductively valid. This is a curious consequence
indeed of the very narrow way we have defined deductive validity of argu­
ments, and we return to further discussion of the point subsequently.
For now however, it is enough to remember that an argument will be
said to be deductively valid if it is impossible for the premisses to be true
and the conclusion false. All that is required then is that the truth of the
premisses be sufficient for the truth of the conclusion — never mind
whether the premisses are true or consistent, whether the argument is free
of vicious circles, or whether the premisses are connected to the conclusion
in any other way. All that we need for deductive validity is the fail-safe
requirement that we must never be able to go by valid deduction from true
premisses to a false conclusion.

2. Formal Logic
Does deductive argument as a type of reasoning have any place in
managing real-life disputes like the parliamentary debate we studied in the
previous chapter? Yes, it does — for two reasons, mainly. First, deductive
logic can be studied as a formal logic — that is, a logic that identifies certain
forms of argument as universally valid. It follows that deductive logic can
be a reliable way of assuring you that your own or your opponent's argu­
ments are deductively correct. Second, if one of your opponent's arguments
happens to be deductively incorrect, you can use this fact to criticize his
argument, provided you observe some precautions. As we will eventually
see in this book, formal deductive logic has other uses as well. It enables us
to reconstruct arguments that are only partially given. It can do this by
showing us different ways of adding to the given argument that would make
it into a valid argument.
66 INFORMAL FALLACIES

Formal logic is that part of logic that has to do with the form of an
argument. What is the form of an argument? Here is an example. Consider
the pair of arguments.
If Bob is a bachelor then Bob is single.
Bob is a bachelor.
Therefore, Bob is single.
If Professor Cresswell writes on the board, the chalk will squeak.
Professor Cresswell writes on the board.
Therefore, the chalk will squeak.
What these two arguments have in common is the same form. This form is
a common one, long acknowledged as a valid form of argument: if the first
thing then the second, but the first thing, therefore the second thing. The
traditional name of this form of argument is modus ponens.
We get the notion of a common form of the above two arguments by
observing that certain aspects of both of them are the same (constant),
while other aspects are variable. What is common to them is the condi­
tional, 'If , then .' What varies is the propositions that go in the
blanks (the first thing and second thing respectively). If we let the first
proposition be symbolized by the variable letter p, and the second be sym­
bolized by the variable letter q, then both arguments have the same form:
if ρ then q, p, therefore q. The 'if ρ then q' and the ρ are the premisses, and
the q is the conclusion.
Below are displayed some deductively valid forms of argument that we
familiarly recognize as forms of reasoning an argument may take.
If ρ then q. If ρ then q. If ρ then q.
ρ Not q. If q then r.
Therefore, q. Therefore, not p. Therefore, if ρ then r.
Modus Ponens Modus Tollens Hypothetical Chain
Here are some examples of arguments that have the respective forms
above.
Modus Ponens
If the kettle is boiling, the water is hot.
The kettle is boiling.
Therefore, the water is hot.
Modus Tollens
If you graduated, you paid your tuition fee.
 THE LOGIC OF PROPOSITIONS 67

You did not pay your tuition fee.

Therefore, you did not graduate.
Hypothetical Chain
If Fred moved his knight, Bob is checkmated.
If Bob is checkmated, Fred won the game.
Therefore, if Fred moved his knight, Fred won the game.
That these forms of argument are universally valid follow from three pre­
sumptions. First, validity in an argument is defined as never going in any
instance from true premisses to a false conclusion. Second, 'If ρ then q' is
defined in such a way that if it is true, you can never go from a true ρ to a
false q. Both validity of arguments and truth of 'if-then' statements are
"truth-preserving" relationships. Third negation, not p, is defined as being
the opposite in just this sense: not ρ is true (false) where ρ is false (true).
Why do we make these presumptions? Well, we don't have to, but they
have always seemed so reasonable that they have become conventions of
what is now called the "classical" approach to formal logic. Perhaps they
just reflect the basic idea of deductive logic as a form of argument that,
whatever else it accomplishes, should never take an arguer by valid argu­
ment from truth to falsehood. That is, deductive logic should always be reli­
ably truth-preserving.
Here we make a distinction between the validity of an argument and
the truth of a proposition. Propositions are true or false. For example, sup­
pose I am holding a piece of chalk in my hand. The statement This chalk is
white' may then be true — if, in fact the chalk is white. But we may also
speak of conditional statements or if-then statements as true or false. For
example, suppose the piece of chalk is long and spindly. Then it may well
turn out to be the case that the following conditional statement is true: "If
I throw this chalk against the board, the chalk will break." But then again,
it might turn out to be false. In any event, we can see that complex state­
ments like conditional statements can take on the properties of being true
or false.
But arguments are not true or false — they are correct or incorrect.
And when speaking of deductive arguments, we say that an argument is
valid or invalid. Arguments are the sort of things that have premisses and
conclusions. And, to repeat, a deductively valid argument is one that never
takes you from true premisses to a false conclusion.
To see why our three forms of argument above must be universally
68 INFORMAL FALLACIES

valid by the criteria set out above, we can reason as follows. Take modus
ponens first. The first premiss says 'If ρ then q.' That is, it tells you that you
can never go from a true ρ to a false q if this conditional is true. So assume
it's true, and that the second premiss, p, is also true. Then q could not be
false. Hence according to our understanding of negation, as agreed above,
q must be true. Hence modus ponens must be valid — on the assumption
that the premisses are both true, the conclusion must be true. You can't
wiggle out of it.
As an exercise, you can go through a similar proof to assure yourself
that modus tollens is universally valid as a form of argument. Let's look at
the third form, hypothetical chain. Does it always have to be valid as a form
of argument? We can prove it as follows.
The first premiss says that you can never go from a true ρ to a false q.
So if we assume that ρ is true then q has to be true, as in our reasoning for
modus ponens above. By similar reasoning, looking at the second premiss
we see that if q is true, r has to be true. Putting these two truths together,
we see that if ρ is true, then given these premisses, r always has to be true.
Hence hypothetical chain is universally valid as a form of argument.
In short, we can see that once we have defined our constants, like 'if ...
then' and 'not' in a certain way, these definitions can determine certain
forms of argument as generally valid or invalid. That is the way of formal
logic.

3. Classical Propositional Calculus

A very simple model of a class of deductively valid arguments that
nevertheless admits of a decisive and clear test of validity and invalidity of
arguments is the so-called classical propositional calculus (PC). The vari­
ables in PC are proposition-letters, p, q, r, ..., and the constants are defined
as follows.

The negation of a proposition ρ has the opposite truth-value of p. ρ  q

(read 'ρ and q') is only true if both propositions are true, ρ v q (read 'ρ or
 THE LOGIC OF PROPOSITIONS 69

q') is true if either or both of the propositions ρ and q are true. Sometimes
another type of 'or' is also introduced called the exclusive disjunction:
ρ q is true if one or the other but not both ρ and q are true. However, this
exclusive type of disjunction can easily be defined as
The way the three constants are defined above, each of them is truth-
functional, meaning that each row has a definite truth-value filled in. This
means that there are no gaps. As soon as we know the value of the basic
propositions ρ and q, the value of the complex proposition made up by
means of the constants and variables is always known. Consider the com­
plex proposition that defines the exclusive disjunction for example.

Ρ q
τ τ Τ F Τ F
τ F F Τ Τ Τ
F τ F Τ Τ Τ
F F F Τ F F
To see how this works, we first of all put down a column of truth-values for
all combinations of truth and falsity of ρ and q. There are four possibilities.
Then we set down the column for ρ  q. The negation ](p  q) will be just
the opposite of the preceding column. Then we set down a column for ρ ν
q. Finally, the whole expression (p v q)  1(  q) is simply the conjunc­
tion of the two previous columns (the order doesn't matter). So we see that
any expression made up of constants and variables, no matter how com­
plex, always has a fixed truth-value provided the truth-values of the compo­
nent propositional variables are fixed. Hence PC is often called truth-func­
tional logic.
Looking at the truth-table above, we can see how the expression (ρ ν
q)  ](  q) is indeed equivalent to the exclusive 'or' meaning 'either ρ or
q but not both.' For if we wanted a truth-table definition of the exclusive
'or,' it would look like this.
p q
τ τ F
T F Τ
F T Τ Exclusive Disjunction
F F F
The exclusive disjunction says that either ρ or q is true, so it is the same as
the truth-table for the inclusive 'or,' except that it says "not both." Hence
70 INFORMAL FALLACIES

the only difference between it and the inclusive 'or' is that the first row,
where ρ and q are both true, will be false for the exclusive 'or.' Notice that
this result is the very same set of truth-values as those in the final column of
the truth-table for Thus can be defined in terms of
and 1.
One other major constant needed for PC is the conditional, 'if ρ then
q.' Are conditionals truth-functional?
Ρ q If ρ then q
τ τ T
τ F F Example: If I drop this
F τ ? chalk, it will break.
F F ?

Suppose I have a long, spindly piece of chalk in my hand. Suppose it's true
that I drop it (p) and it's true that it breaks (q). Then most of us would
probably agree that the conditional, 'If ρ then q' is true. Suppose however
that I drop it, but it doesn't break. Then most of us would say that the con­
ditional 'If ρ then q' (If I drop it, the chalk will break) is false. So the first
two rows of the truth-table are truth-functional.
But what are we to say if I don't drop the chalk at all, as in the last two
rows of the table? Most of us would probably say that until you drop the
chalk, you can't really tell whether the conditional 'If I drop the chalk it will
break' is true or false.
The lesson of this example appears to be that 'If ... then' is not truth-
functional. However, in order to have a simple logic, it is desirable to have
all the complex propositions determinable in truth-value only by the truth-
values of the basic propositions. Can we legislate a truth-functional 'If ...
then' and still get a useful logic. How to do it?
If we let both question-marks be F's, we get an 'If ... then' that is the
same as 'and.' But that is no good. For 'If ... then' is very different from
'and.' If we let the first question-mark be Τ and the second F we get a col­
umn of truth-values the same as that for q. But that can't be right either. 'If
I drop the chalk it will shatter' is very different from 'The chalk will shatter
(whether I drop it or not).' As a third trial, what about putting F for the
first question-mark and Τ for the second? This result would say that 'If ρ
then q' only comes out true when ρ and q have the same truth-value. By
this approach 'If ρ then q' would always be equivalent to 'If q then p.' It
wouldn't matter which way you go. But 'If ... then' is not like this. In gen-
 THE LOGIC OF PROPOSITIONS 71

eral, 'If I drop this chalk it will shatter', and 'If this chalk will shatter, I drop
it.' are quite different. The second seems to imply that the only way it could
shatter is if I dropped it. Whereas the first has no such implication. And in
general, 'If ρ then q' is not the same as 'If q then p.' 'If I graduate, I will
have paid my tuition' is quite different in its truth-conditions from 'If I pay
may tuition, I will graduate.'
Of all the possibilities, only one remains.

Ρ q p⊃q
τ τ τ
τ F F
F τ τ Material Conditional
F F τ
The type of conditional defined in the above table is called the material con­
ditional or "hook." In keeping with the general account of validity for
deductive logic, it only comes out false if the ρ is true and the q is false.
Otherwise it is ruled to be true. This definition of the conditional, while
admittedly arbitrary in the last two rows, is truth-functional and still pre­
serves many desirable valid forms of deductive argument. Below are some
valid deductive forms of argument in classical PC.
Modus Example: If the Rangers win, the
Fonens Bruins are out of the playoffs. The
q Rangers win. Therefore the Bruins
are out of the playoffs.
Modus Example: If the Rangers win, the
Tollens Bruins are out of the playoffs. The
Bruins are not out of the playoffs.
Therefore, the Rangers do not win.
Transitivity Example: If I take your knight I
Of ⊃ put you in check. If I put you in
check, I win the game. Therefore,
if I take your knight, I win the game.
In general, the above way of defining the conditional in classical PC turns
out to be reasonable as long as all you mean by 'If ρ then q' is 'It's not the
case that ρ is true and q is false.'
The splendid thing about classical PC is that there is always a finite
procedure for determining precisely whether any form of argument expres-
72 INFORMAL FALLACIES

sible in classical PC is valid or invalid. All you need to do is construct a

truth table. Consider modus ponens.
Ρ q ⊃q
(1) τ τ τ
(2) τ F F
(3) F Τ Τ
(4) F F τ
P q Modus
Ρ Ponens
q
As you scan over the four rows of the truth-table, you easily see that there
is no row where both premisses ( p D q and p) are true and where the con­
clusion (q) is false. Hence modus ponens must be valid. However consider
this form of argument: ρ ⊃ q , q; therefore p. Sometimes called affirming
the consequent, this form of argument is easily shown to be invalid by look­
ing at the same truth table above. In row (3), ρ ⊃ q is true and q is true, but
ρ is false. Hence this form of argument must be invalid — it permits a pos­
sible assignment of truth-values that takes us from true premisses to a false
conclusion. And we remember that a deductively valid argument is one that
can never take us from true premisses to a false conclusion.
No matter how many component propositions we begin with, classical
PC always provides a test for validity or invalidity. One further example
may serve to illustrate.
Example: If Bond goes to Istanbul
he will be attacked by mad scien-
Dilemma tists. If Bond goes to Macao he
will be followed by KGB assassins.
Either Bond goes to Istanbul or
Macao. Therefore, either Bond
will be attacked by mad scientists
or followed by KGB assassins.
The truth-table for this form of argument is quite long, since you have to
have 2 4 = 16 rows of a truth table in order to represent all possible combi­
nations of truth-values for p, q, r and s. However, there is a short-cut
method called the fell swoop (Quine). We want to see if there is a possible
assignment of truth-values that could make all three premisses true and the
 THE LOGIC OF PROPOSITIONS 73

conclusion false. So we could reason as follows, to see if this is possible.

In order for the conclusion (q v s) to be false, both q and s have to be
false. Reason: q v s is only false if both disjuncts are false, by definition.
Assuming the conclusion is false, could all the premisses be true? Well, if q
is false, the first premiss (p ⊃ q), could only be true if ρ is false. Reason:
otherwise ρ ⊃ q would have the values T ⊃ F and thereby be false, by the
truth-table definition of ⊃. Similarly, the second premiss can only be true if
r is false, given that we know that s has to be false for the conclusion of the
whole argument to be false. However if both ρ and r are false, as is required
to make the first two premisses true, then the third premiss (ρ ν r) has to
be false. In other words, the only possible assignment of truth-values that
makes the conclusion false is such that if the first two premisses are to be
true, the remaining premiss has to be false. In short, if the conclusion is
false, the premisses can't possibly all be true! Consequently the argument
must be valid. Reason: it is just not possible to assign truth-values in such a
way that the premisses are all true while the conclusion is false.
If you have trouble following all the reasoning of the paragraph above,
construct a truth-table yourself for the dilemma and carefully scan across all
the rows of the table. If you have constructed it right, there will be no row
of the table (possible assignment of truth-values to the basic propositions)
where the premisses are all true and the conclusion is false.
Now we can see the beauty of formal logic. Once you can identify the
form of an argument, you can always decide whether a given argument is
valid or not.

4. Applying Deductive Logic to Arguments

Now we have seen some examples of deductively valid forms of argu­
ment. You may want to examine some deductively invalid ones as well.
These are ones where, in some instances, you can go from true premisses to
a false conclusion. Two examples are given below.
If ρ then q. If ρ then q.
q Not p.
Therefore, p. Therefore, not q.
Affirming the Denying the Antecedent
Consequent
An example of the former would be: "If you graduated, you paid your tui­
tion fee. You paid your tuition fee. Therefore, you graduated." Clearly this
74 INFORMAL FALLACIES

is an invalid form of reasoning. In the example, you may have paid your tui­
tion fee, but failed all your courses. Hence, it does not follow that you have
graduated. Look at the premisses of affirming the consequent. The first one
says that you'll never go from a true ρ to a false q. But that doesn't mean
that you couldn't go from a true q to a false p. Hence, if the second premiss
is true, that doesn't exclude the possibility of the conclusion being false.
For similar reasons, denying the antecedent turns out to be invalid as a
form of deductive argument. For example, the following argument is incor­
rect: "If you're in Toronto then you're in Canada. You're not in Toronto.
Therefore, you are in Canada." This argument is invalid because the pre­
misses could be true in the case where you're in Chicago. But it does not
follow that you are in Canada.
Now we can see the promise of deductive logic as a useful tool in
evaluating arguments. If we can clearly differentiate between valid and
invalid forms of argument, we are now surely in a position to get the best of
any opponent who advances invalid arguments or refuses to be convinced
by our own valid ones.
Perhaps unfortunately, applying deductive logic is not so straightfor­
ward as one might think or like. First, an opponent who does not wish to
accept the conclusion of your deductively valid argument can simply reject
the premisses of it. We have already noted that a deductively valid argu­
ment is one that has a form that can never go from true premisses to a false
conclusion. But it does not follow that the same deductively valid argument
cannot go from a false conclusion to false premisses. Quite possibly it
could.
The lesson in general is this. Your opponent must accept your conclu­
sion if your argument has a deductively valid form and the premisses are
accepted as true by that opponent. Otherwise there is always the loophole:
"I accept the validity of your argument. But its very validity, coupled with
the fact that I cannot accept your conclusion, forces me to infer that I can­
not accept all of your premisses." In such a case, your use of deductive logic
might not only fail to force the concession of your conclusion by your oppo­
nent in argument. It might even point out to him which group of your pre­
misses he needs to argue against in order to refute your argument. It seems
then that one needs to be careful in how one might attempt to use deductive
logic as a method of reasonably convincing an opponent or making a case to
refute that opponent's position. However, deductive logic could be useful
in locating more exactly the set of propositions on which participants in
 THE LOGIC OF PROPOSITIONS 75

argument disagree.
What then is the value of formal deductive logic in argument? Its value
is that a valid form of deductive argument is universally valid in the sense
that every argument that is an instance of it must never have both true pre­
misses and a false conclusion. This means that if your opponent rejects the
validity of a valid argument he may, with all the force of deductive logic, be
accused of being inconsistent. And logical inconsistency of position in dis­
pute is the most deadly form of vulnerability. We have seen the power of
allegations of inconsistency in the examples of ad hominem criticism we
studied in the previous two chapters. We will subsequently see other exam­
ples.
One who is clearly and provably logically inconsistent in the position
he advocates often assumes the burden of seeming to be either a fool or a
hypocrite. Persistence in an inconsistent position can easily make an arguer
open to ridicule, and perhaps even eventually to accusations of insanity or
incompetence to understand rational discourse. Needless to say, this is not
a good position for an arguer to be in. And sometimes it is not an easy posi­
tion to get out of.
What we are saying is that deductive formal logic can be used in argu­
ment as a powerful vehicle for criticism of an opponent's position. Suppose,
for example, your argument can be shown to have the form of modus
ponens. And suppose your opponent is foolish enough to reject your argu­
ment as incorrect. Suppose in fact he accepts the premisses of it, even
though he rejects the conclusion. What do we have here?
Your opponent, let's say, rejects the following argument.
If a million people are unemployed, Canada is in a recession.
A million people are unemployed.
Therefore, Canada is in a recession.
He rejects the conclusion, but he is not inclined to reject the premisses. In
fact, let's say, it can be shown from his conceded position as a whole that he
is committed to accepting these premisses.
Now we know from deductive logic that the above argument has the
form of a valid argument. Consequently, the above argument is, without
question deductively valid. Just repeating that the argument is valid is not
likely to impress your opponent or your audience, unless they too know
about and accept classical formal logic. How could you use formal logic to
criticize your opponent's vulnerable position in this instance?
76 INFORMAL FALLACIES

The way to do it is as follows. Your opponent is inclined to accept the

premisses. So you ask him: "Do you agree that a million people are
unemployed?" He might reply: "I don't deny it." Then you ask him about
the other premiss: "Do you accept the proposition that if a million people
are unemployed, we are in a recession?" Again, he is not likely to agree,
but let's say he concedes that this is possible by replying: "I don't deny it."
How you criticize his position is by way of modus tollens. First, by
means of examples, or any other means of persuasion, you get him or your
audience to agree that, whatever else is true, modus tollens is a valid form
of inference. Having secured agreement on this, you then argue as follows.
"You state firmly that Canada is not in a recession. But suppose it's true
that if a million people are unemployed, Canada is in a recession. Put these
two statements together, and it can't be true that a million people are
unemployed. So you must reject the one statement or the other, or you are
inconsistent. You must choose one or the other of the pair below.
Canada is not in If a million people are
a recession. unemployed, Canada is
in a recession.
But you don't deny the first one, by your own acknowledgement. And
given that you don't deny the first one, you can't accept the second one
unless you also accept that it's false that a million people are unemployed.
So deductive logic assures us. But as you already admitted, you don't deny
that a million people are unemployed. Therefore, you are inconsistent and
your position is incoherent and worthless. You accept the very same propo­
sition that you deny."
Here then is an instance revealing how deductive logic can be useful in
criticism. Any participant in argument who is ambivalent in what he accepts
or refuses to accept as part of his position can be criticized by reducing his
position to inconsistency using deductive logic.
Moreover, deductive logic can also be used as a positive tool for build­
ing up your own position in argument. 1 For remember that any argument
that has a valid form can never take you from true premisses to a false con­
clusion. Consequently, if your premisses are acceptable as true, any conclu­
sion you derive from them by valid inference must also be acceptable as
true.
Notice however that we have been discussing the use of deductive logic
by one participant in argument to criticize the position of another partici-
 THE LOGIC OF PROPOSITIONS 77

pant in the argument. If a participant's position is ambivalent, 2 then an

arguer can, by means of dialogue, carry through an argument to reduce his
position to inconsistency. But deductive logic is only one part of this whole
process of refutation by dialogue.. All the other elements of dialectic we
covered in chapter 1 must clearly also play key parts in the whole process of
criticism or refutation thereby carried out.
Consequently our thesis is that classical deductive logic should be con­
tained in the concept of logical dialogue to give us a theory of argument
adequate to understanding how criticism of arguments can be reasonable
and justifiable. Propositional logic, so conceived, is the inner core of argu­
ment. The game of dialogue is the other shell of argument. The two com­
bined offer a theory of argument that shows how logic can be applied to
realistic argumentation. The remainder of this monograph will be devoted
to arguing towards the thesis that the best way to understand the normative
force of reasonable criticisms and well-argued refutations in realistic con­
troversies is through the use of formal logic in dialogue. But these days, one
must also ask "Which formal logic?"
There are many problems inherent in our argument, and these prob­
lems will continue to subject to controversy and further study as the field of
argument study develops. One sort of problem is posed by the fact that
there are now many other formal, deductive logics other than classical
logic. Hence we will have to study questions of which non-classical logics
might be appropriate for various contexts of dialogue. There are also many
theoretical questions of the relationship between semantics and pragmatics
in the study of argument. If classical logic represents "semantics" and regu­
lated games of dialogue represent "pragmatics," then it seems that, accord­
ing to the proposal above, semantics is (in a way we will see) contained
within pragmatics. However, there are many controversies about the nature
of this sort of relationship. We will return to these controversies sub­
sequently. For the present, let us at least provisionally concede that classi­
cal deductive logic can be usefully applied to argumentation within at least
some contexts of dialogue. If we are right that this presumption will turn
out to be reasonable, then applied logic is possible within the present state
of the art of logic as a discipline.

5. Invalidity and Fallaciousness

Buoyed up by the prospects of applying classical logic as a powerful
tool, we might be inclined to think that an opponent in argument could also
78 INFORMAL FALLACIES

be criticized for committing formal fallacies. We might think then that if an

opponent's argument can be shown to have the form of an invalid argument
like affirming the consequent, his position can be refuted as worthless. But
there are grounds for caution in this negative use of formal logic.
It so happens that not all instances of invalid forms of argument are
themselves invalid arguments. An example will illustrate.
If California is in the U.S., then California is in the U.S. and Los
Angeles is in the U.S.
California is in the U.S. and Los Angeles is in the U.S.
Therefore, California is in the U.S.
This argument is of course valid, for it is impossible for the premisses to be
true and the conclusion false. If even the second premiss is true, the conclu­
sion has to be true as well. Yet this argument has an invalid form. Let q
stand for 'California is in the U.S. and Los Angeles is in the U.S.' and ρ
stand for 'California is in the U.S.' Then our argument above has the form
of affirming the consequent, an invalid form: if ρ then q; q; therefore p.
Here then we have an invalid form of argument, but an instance of it hap­
pens to be valid.
This peculiar consequence stems from the presumption that the
schematic letters ρ and q can stand for any propositions whatever. The only
restriction is that once you have used a letter, say p, for a proposition, then
you must always uniformly use the same letter, p, to stand for that proposi­
tion everywhere else in your representation of the form of the argument.
That rule of uniformity was certainly adhered to in our representation of
the form of the argument above.
The problem however is posed by the fact that the same argument can
have various different forms. For example, suppose we let ρ stand for
'California is in the U.S.' and q stand for 'Los Angeles is in the U.S.' Then
our representation of the form of the argument above will be this: if p, then
ρ and q; p and q; therefore p. This form is certainly valid. For even if the
second premiss is true, the conclusion has to be true.
One might criticize the argument above, having seen our second rep­
resentation of its logical form, by pointing out that the first premiss is "re­
dundant" or "irrelevant." The first premiss is not even needed to establish
validity. For if the second premiss is true, the conclusion cannot possibly be
false. Indeed, one might also criticize the argument as being circular or tri­
vial. In effect, all it says is this: "California is in the U.S. and Los Angeles
 THE LOGIC OF PROPOSITIONS 79

is in the U.S., therefore California is in the U.S." If asked to prove that

California is in the U.S. and I offered that argument to prove it, you could
quite justifiably criticize me for arguing in a circle! These observations are
quite revealing about the limitations of deductive logic. For neither circu­
larity nor irrelevance in argument make any difference to, or are modelled
by classical deductive logic.
But back to the main point, namely this. The same argument can have
various different legitimate representations of its logical form. Sometimes
the same argument might have one valid form and one invalid form! The
following lesson must be our conclusion. If it is shown that an argument has
a deductively invalid form, it need not follow that the argument is invalid.
It follows then that deductive formal logic is more limited as a tool of
negative criticism for the refutation of an opponent's argument by finding
formal fallacies exemplified in his argument. The formal invalidity does not
necessarily imply that the argument itself is invalid.3
The problem here is that there are degrees to which any formal rep­
resentation of an argument cuts into the deeper structure of that argument.
In our example above, it could be said that the second formalization was
more complete or deeper than the first. But there is no general way for tel­
ling us when a formalization of an argument is deep enough to include all
the forms it could possibly have. Consequently, deductive formal logic can
never by itself be a reliable test for invalidity or incorrectness of sequences
of argument in a natural language like English.
What can we say then about formal fallacies? It seems that the best we
can say is that if an argument is formally invalid, it may be open to criticism
as falling short of deductive validity. But then again, it may not be, either.
We can only refute an opponent's argument in the weak sense of refutation
— we can say that he hasn't established his conclusion yet deductively by
means of formal logic. But it would be a fallacy to conclude that his argu­
ment is refuted in the strong sense — we can't say that it has been shown to
be invalid.

6. Relevance and Validity

The decisive advantage of formal deductive logic over the unregulated
quarrel and debate as a method of evaluating the correctness of an argu­
ment lies in the fact that formal logic evaluates certain objective relation­
ships in sets of propositions. Thus formal logic appeals to objective factors
to settle disagreements rather than leaving the issue to the sayso of the dis-
80 INFORMAL FALLACIES

putants themselves, for they may never agree on whose argument is better
if left to their own devices.
On the other hand, formal logic has certain limitations built into its
structure. We remember that in classical formal logic, an argument is valid
if it is not possible for the premisses to be true and the conclusion false.
This rather narrow definition of validity ignores many factors that one
might wish to take into account in evaluating arguments however. Foremost
among these is the factor of relevance.
Is classical deductive logic a good model of natural argumentation
where disputes that may involve fallacies take place? This question is itself
a disputed one and has to be approached with some caution. The main
objection to classical logic as such a model concerns certain inferences that
have a valid form in classical logic but that do not seem to represent correct
arguments. The two most famous of these "paradoxical" inferences are
these.
Example: It is not raining. Therefore,
if it is raining then 2 is a prime number.
Example: Auckland is in New Zealand.
Therefore, if chlorine is heavier than
air then Auckland is in New Zealand.
As the examples suggest, this sort of reasoning seems bizarre or astound­
ing. But the forms of argument on the left are undoubtedly valid in classical
deductive logic. Consider the top one. The only way the conclusion Tf p
then q' could be false is if ρ is true and q false. But if ρ is true, the premiss,
not p, must be false. Hence there is no consistent assignment of truth-val­
ues that could make the premiss true and the conclusion false. Therefore,
by the classical account of validity, any argument of that form must be
deductively valid.
These arguments are certainly correct from the classical point of view
— they could never take us from true premisses to a false conclusion. But
what is it about them that makes them seem deleriously inappropriate as
correct arguments?
Well, certainly one thing about them is that the basic propositions, the
ρ and q, don't seem to have any connection with each other. What does the
weather have to do with 2 being a prime number or not? What does Auck­
land have to do with the weight of chlorine? We were most concerned
about fallacies of relevance in chapter 1. If there are such things as fallacies
 THE LOGIC OF PROPOSITIONS 81

of relevance, these argument forms must be the granddaddies of them all!

Can an argument be valid yet fallacious? We seem caught in a conundrum.
However, perhaps at this point it is well to remember that classical
logic did not really address itself to the notion of relevance. We purposely
defined 'If ρ then q' in the previous section in such a way that all that was
demanded that we never go from a true ρ to a false q. The 'If ... then' in
classical logic is defined in a manner in keeping with the conception of val­
idity there adopted. Classical logic does its job as long as it never allows us
to go by valid argument from true premisses to a false conclusion — never
mind the intermediate steps or interconnections of how we got there.
Perhaps then, classical logic is not wrong as an account of correct argument,
but it is simply limited to one aspect of the nature of correct argumentation.
Relevance is yet another aspect.
One approach is that of Grice (1975) who argues that classical logic is
correct but incomplete in that it requires supplementation by conversa­
tional trimmings in order to fully reflect the argumentation of natural con­
versational interchanges. According to Grice, we normally follow co-opera­
tive principles of conversation like the maxim, 'Be relevant!' By the conver­
sational approach then, the above two inference-forms are not incorrect but
simply incomplete. In polite conversations, we avoid the rudeness of
irrelevant transitions like 'If it's raining then 2 is a prime number.' But from
a point of view of deductive logic, there's nothing wrong with such a condi­
tional, the question of relevance apart.
One problem with Grice's approach however is that relevance becomes
a matter of conversational politeness rather than a matter of precise logical
regulation. Consequently, if there arises a dispute about whether two prop­
ositions in an argument are relevant or not, it is not clear how it is to be set­
tled. Grice offers no precise guidelines. So it seems we are back at the level
of the quarrel and the debate. If we can't agree whether ρ is relevant to q
or not, and the dispute turns on the issue of relevance, how is it to be set­
tled? Politeness may help, but may not resolve a substantive dispute. It
seems we are back to the psychological criterion of adversarial rhetoric.
Whoever can persuade the opposition, the audience, or the referee to his
side — that ρ is relevant or irrelevant — wins the argument. But is this
approach good enough?
Much depends on how seriously we take the claim that failure of rele­
vance really is fallacious. Is the irrelevant argument truly incorrect, or is it
merely a lapse of manners or rhetorical persuasion?
82 INFORMAL FALLACIES

7. Subject-Matter Relatedness
One recent approach takes seriously the thesis that irrelevance is a log­
ical failure. This approach, initially suggested in a seminar in 1976 by David
Lewis, postulates that an argument can be thought of as taking place rela­
tive to a set of topics. Let us call this set of topics, T, the most specific set
of topics that represent what the argument is about. In many quarrels and
debates, the set Τ is never clearly specified, but that does not mean that it
couldn't be, or shouldn't be! If relevance is at issue, the set Τ should always
be carefully specified in advance by the disputants. Then what we do is to
assign to each basic proposition in the argument a subset of Τ called the
subject-matters. The subject-matters represent the topical content of each
basic proposition in the argument.
As a simple illustration, supposing the set Τ in an argument is
{bananas, yellow, nutritious, edible}. And suppose we encounter two prop­
ositions: ρ is the proposition 'Bananas are yellow' and q is the proposition
'Bananas are nutritious.' Then the subject-matter of ρ is the set {bananas,
yellow} and the subject-matter of q is the set {bananas, nutritious}. In
other words, each proposition in the argument will take on not only a set of
truth-values, as in classical logic, but also a set of subject-matters. Then it
is natural to rule that ρ is relevant to q in one important sense (or better, ρ
is related to q) if ρ shares subject-matter overlap with q. In the present
example, we say that 'Bananas are yellow' is related to 'Bananas are nutriti­
ous' because each proposition shares the common topic 'bananas.' How­
ever, in the two "paradoxical" argument forms above, it is easy to see a fail­
ure of relatedness. For example, 'It is raining' does not share any common
subject-matters with '2 is a prime number.'
Relatedness of propositions in argument could refer to many different
types of relationships. Clearly however, one fundamental type of related­
ness is subject-matter overlap of propositions. As a general notion, subject-
matter relatedness has three defining properties. First, it is a reflexive rela­
tion — that is, a proposition is always related to itself. For example
'Bananas are yellow' shares subject-matters with itself. Second, it is a sym­
metrical relation — that is, if ρ is related to q then q must be related to p.
For example, if 'Bananas are yellow' shares subject-matter with 'Bananas
are nutritious' then the second proposition must also share a subject-matter
with the first. One could scarcely doubt that subject-matter relatedness has
these two properties.
But when we come to a third property, that of transitivity, we see it
 THE LOGIC OF PROPOSITIONS 83

fails. It may be the case that ρ is related to q, and q is related to r, yet it may
not be true that ρ is related to r. For example, 'Bananas are yellow' is
related to 'Bob ate six bananas' and 'Bob ate six bananas' is related to 'Bob
has six children.' However, 'Bananas are yellow' does not share any sub­
ject-matters with 'Bob has six children.' Thus subject-matter relatedness is
not transitive, as a general characteristic. 4
Now we have at least one clear basic idea of what relevance in argu­
ments could mean, how does such a conception affect the correctness or
incorrectness of arguments? Following a formalization of Epstein (1979) we
can construct a relatedness propositional logic that will be just as formal a
logic as classical logic.
The 'if ... then' (conditional) is a fundamental way of forming complex
propositions in any formal logic. In classical deductive logic, 'If ρ then q' is
defined so that it is false only where ρ is true and q is false. This approach
reflects the basic idea of classical logic that hypothetical reasoning should at
least be truth-preserving, whatever its other properties might be. Hence 'p
⊃ q' (if ρ then q) in classical logic is defined as true except where ρ is true
and q is false. In relatedness logic however, the 'if ... then' incorporates the
idea that ρ and q share some common subject-matter (are related to each
other). Hence, the conditional in relatedness logic, ρ→ q, is defined as true
except where ρ is true and q is false, or where ρ and q are not related. The
negation of p, lp, is defined the same way in both logics. It is false where ρ
is true, and true where ρ is false.
This new type of formal logic does take into account whether the sub­
ject-matters of propositions are related to each other in an argument. So it
does at least partially overcome the failure in classical logic to take rele­
vance into account in evaluating arguments. To see how, let us go back to
consider the examples above of "paradoxical inferences."
The reason why the first one seemed bizarre was that 'It is raining'
seemed to have nothing to do with '2 is a prime number.' This failure is
reflected in the way relatedness logic deals with the argument. Since 'It is
raining' shares no common subject-matters with '2 is a prime number,' the
inference fails to be valid. Reason: the conclusion 'If it is raining then 2 is a
prime number' is false, because of failure of subject-matter overlap of the
two component propositions it contains. Yet the premiss, 'It is not raining'
could quite well be true even while this conclusion if false. Hence the argu­
ment is not valid in relatedness logic. Similar considerations show the
invalidity of the second inference in relatedness logic.5
84 INFORMAL FALLACIES

Now we have at least a working idea of one thing that might be meant
by "relevance." But in subsequent chapters, we will see that there are other
kinds of "relevance" as well as subject-matter overlap that play important
roles in the study of fallacies. With relatedness we have just begun the study
of relevance in argumentation. Subject-matter relatedness turns out to be
one important aspect of relevance that plays a role in the analysis of argu­
ments, but it is only one among many aspects.
Next, we show how Epstein (1979) has developed a formal proposi-
tional calculus based on the idea of relatedness.

8. Relatedness Logic
First we have to define new constants, in line with our new model of
argument. The most trouble we had in classical logic was with the condi­
tional. The problem seemed to be that completely unrelated propositions
could make up true conditions just because of their individual truth-values.
This problem suggests that in order for a relatedness conditional to be true,
the basic component propositions should be related by common subject-
matters. The following type of definition is thereby suggested. Let (p,q)
stand for 'p is related to q.' What we mean by (p,q) can vary for different
interpretations. However, in line with the present discussion, we mean by
this relation that ρ and q share some common subject-matters.
In this new kind of formal logic, we need to define our logical constants
in a new way. The most important constant is 'If ... then.' How do we
define it, to take into account subject-matter relatedness? The best
approach is to define a new conditional constant → that is the same as the
classical ⊃ except that it comes out false if ρ is not related to q. The truth-
table for → is below.
Ρ q p→q
τ τ Τ τ Relatedness Conditional:
τ F Τ F ρ → q is false only if
F τ Τ Τ (1) it is the case that
F F Τ Τ ρ is true and q is false
Τ τ F F or (2) ρ is not related
Τ F F F to q.
F τ F F
F F F F
 THE LOGIC OF PROPOSITIONS 85

By the above definition, ρ → q requires both that the truth-values are right
(like the material conditional) and also that ρ is related to q.
Now how do we define the other connectives? First consider negation.
Here relatedness does not seem to matter. If ρ is related to q, then lp will
also be related to q. If ρ is not related to q, then ]p will not be related to q
either. For example, if 'Bananas are yellow' is related to 'Bob ate a
banana,' then 'Bananas are yellow' will also be related to 'Bob did not eat
a banana.' Hence negation can be defined the same way as in classical logic
— we need not worry about the factor of relatedness at all.
Similarly, with conjunction relatedness does not seem to be important.
If we conjoin together two propositions 'Bananas are yellow' and 'Bob has
six children' then that conjunction is true if both of the component proposi­
tions are true, never mind that one is not related to the other. Hence we
can define conjunction using the same truth-table as in classical PC.
Disjunction, however, appears to require relatedness. That is, the
proposition 'Bananas are yellow or 2 is a prime number' does not seem to
be true because the two component propositions are unrelated. Therefore
we define the relatedness 'or' as follows: 'p v q' is true just in case (1) at
least one of the pair {p,q} is true and (2) ρ is related to q.6
Now we have given truth-tables for all the constants of relatedness PC,
we seem to have all we need for a logic. But there is still one question to be
resolved. In classical PC all the constants were truth-functional. But these
new ones, at least the —» and v, are not. Thus we have to make a decision
on how the basic propositions are related to the complex ones. When is a
simple proposition ρ related to a conditional, q →r? Does ρ have to be
related to both q and r, or is it enough that it be related to just one of them?
For example, is 'Bananas are yellow' related to 'If Bob is a canary then Bob
is yellow'? Well, it does seem to be. That is, 'Bananas are yellow' is related
to part of the conditional 'If Bob is a canary then Bob is yellow' (namely the
consequent), so it seems that we want to say it must therefore share some
subject-matter with the whole conditional. If so, the rule we need to adopt
is this one: ρ is related to q → r just in case ρ is related to q or ρ is related
to r.
We adopt similar rules for  and ν: ρ is related to q  r(q v r) just in
case ρ is related to q or ρ is related to r. Thus the general approach is this:
one complex proposition is related to another complex proposition if any one
component proposition of one complex is related to any one component of
86 INFORMAL FALLACIES

the other complex. For example if we have two complex propositions '(p →
q) v ( ^ Is' and '(t v u) → lw' then we know that they have to be related
if any of their parts are related. Suppose for example that q is related to w.
Then we immediately know that the two complex propositions are related
to each other.
Just as in classical PC, there is always a finite decision procedure for
determining correctness or incorrectness of arguments in relatedness PC.
Consider modus ponens in relatedness logic.
Ρ q p→q
(1) Τ τ Τ τ
(2) τ F Τ F
(3) F τ Τ τ
(4) F F Τ τ
(5) Τ Τ F F
(6) Τ F F F
(7) F Τ F F
(8) F F F F

As you scan over the eight rows of the truth-table, you can see that there is
no row where both premisses (p → q and p) are true and where the conclu­
sion (q) is false. Hence modus ponens is valid in relatedness PC.
Also, just as in classical PC, we can see that affirming the consequent
is invalid in relatedness PC. In row (3), ρ → q is true and q is true, but ρ is
false.
Hence we see that the conception of validity is the same in relatedness
logic as in classical logic. A valid argument is one that never goes from true
premisses to a false conclusion. But the difference is that in relatedness
logic, subject-matter relevance is explicitly taken into account in the way
we define the constants of the logic.
We should also note that relatedness PC will turn out to be a subsystem
of classical PC. All arguments valid in relatedness PC will also be valid in
classical PC. But there are some forms of argument that are valid in classi­
cal PC but that fail to be valid in relatedness PC. Some examples may be
instructive.
 THE LOGIC OF PROPOSITIONS 87

Transitivity of the
Relatedness Conditional

Could the premisses of this form of argument be true while the conclusion
is false? Well, we have already seen that consideration of truth-values alone
would not permit an assignment that would make the premisses true and
the conclusion false. But as formulated above, the form of argument with
→ instead of ⊃ requires the right arrangement of subject-matter relation­
ships as well. What if the subject-matter of ρ fails to be related to that of r.
Still, it is quite possible that ρ is related to q, and q is related to r. If the
truth values were right in such a case, then both premisses would be true
and the conclusion false. Thus there is at least one row of the truth-table
where both premisses are true and the conclusion is false. One of them is as
below.

This possible assignment of truth values makes the premisses true and the
conclusion false. Hence the form of argument above for transitivity of the
relatedness conditional is invalid.
Another valid argument form of classical PC that fails to be valid in
relatedness PC is exportation.

Looking on the right, we see that exportation as a form of argument is valid

in classical PC. The only way the conclusion could be false is if ρ and q are
both true and r is false. But given these values, the premiss must be false as
well. The premiss cannot be true while the conclusion is false. Hence expor­
tation is valid in classical PC.
But what about validity in relatedness PC? Look at the form on the
left. Consider these values.

If q is not related to r (as above, then the part q → r of the conclusion must
be false. But if ρ is true (as above), then the whole conclusion ρ → (q → r)
must be false by virtue of the truth-values. However, it is still possible for
88 INFORMAL FALLACIES

the premiss to be true, even given these values. If p,q, and r are all true,
then ( p ^ q ) → r will be true provided one or the other of ρ or q is related
to r. To make the conclusion false, we made q not related to r. But we
could still consistently assume that ρ is related to r (as above). Such an
assignment (as above) makes the premiss true and the conclusion false.
Hence this form of argument is not valid in relatedness PC.
These examples show that relatedness PC is every bit as much of a pre­
cise decision procedure to determine validity or invalidity of arguments as
classical PC. However, the class of valid arguments turns out to be differ­
ent. Reason: in relatedness logic, relevance of subject-matters is taken into
account.
In fact, relatedness propositional calculus turns out to be a subsystem
of classical propositional calculus in just this sense. All the valid forms of
argument in relatedness logic are classically valid forms. But there are some
forms of argument, like the examples we have examined above, that are
valid in classical logic but invalid in relatedness logic.
The exciting thing about relatedness logic is that it offers a formal deci­
sion procedure that takes subject-matter relatedness of propositions into
account. It therefore offers one clear and potentially useful way of defining
relevance in argumentation, and of managing criticisms of arguments on the
basis of claims of irrelevance. This is an exciting prospect.
However, there are grounds for caution. Our studies of fallacies like
the ad hominem, ad populum and so forth, have already suggested that
many objections to arguments on grounds of irrelevance pertain to factors
in argument other than subject-matter relatedness. Many a fallacious ad
hominem argument is effective as a persuasive strategy precisely because
the "irrelevant" personal claim made is indeed subject-matter related to the
conclusion or proposition at issue. Despite subject-matter relatedness, the
argument may be perceived as fallacious.
In fact, many faults of argument perceived as failures of relevance are
pragmatic failures in procedures of proper dialogue, rather than failures of
subject-matter relatedness.

9. Semantics and Pragmatics

A basic theoretical problem for the study of fallacies, enthymemes,
and other aspects of practical argumentation is the relationship between the
pragmatic concepts of criticism, refutation, fallacy, and so forth, and the
semantic concepts of truth and valid argument. The semantics of argument
 THE LOGIC OF PROPOSITIONS 89

has to do with relationships like validity and consistency conceived of in

terms of truth-conditions. A valid argument is one where it is never true in
the semantical model that the premisses are true and the conclusion false.
An inconsistent set of propositions is one where all members of the set can­
not be true in a model.
As far as semantic evaluations are concerned, a valid argument is one
where you can never go in any case from true premisses to a false conclu­
sion. By valid deductive argument, truth is always preserved. In this
framework, it doesn't matter where the premisses came from. They are
simply a designated set. It doesn't matter where the conclusion came from,
strictly speaking, either. For nothing in the semantic model determines
which proposition in the set designated as the argument comes after the
"therefore." Semantically, it doesn't matter whether the premisses are
related to the conclusion — at least in the classical model of argument —
because if you start with an inconsistent set of premisses, then you can
never go from true premisses to false conclusion, even if the conclusion is
entirely unrelated to the premisses. Semantically, it doesn't matter whether
the argument is circular, or whether the premisses are well-established, or
whether the conclusion is related to the issue of some particular dispute.
From a pragmatic point of view, however — the point of view appro­
priate to the study of informal fallacies — a proposition may not be a
proper conclusion of an argument unless it is appropriately related to the
issue under dispute. An arguer's conclusion is, ultimately, his thesis to be
proven in that context of dispute or dialogue. What an arguer may take as
premisses in a proof are therefore limited to certain classes of propositions.
They may, in some contexts of dialogue for example, be restricted to prop­
ositions that are commitments of the other party to the argument. 7
Semantics then is a matter of one kind of relationship between the pre­
misses and conclusion of argument. However, what really defines those
propositions as "premisses" and "conclusion" is a matter of pragmatics.
Pragmatics is concerned with where the premisses came from, and whose
premisses they are in the argument. Pragmatics is concerned with where the
conclusion came from, its relationship to the issue.8 Pragmatics is con­
cerned with certain relationships between premisses and conclusion, com­
prising the general direction and flow of the over-all argument as a chain of
sub-arguments in a sequence of replies to questions.
At present, there is considerable disagreement about the relationship
between semantics and pragmatics. Many of those in "informal logic" feel
90 INFORMAL FALLACIES

that semantics has no place in the reasonable evaluation of real arguments.

The more traditionalist defenders of formal logic as the only important
model of argument take the opposite view that semantics is all there is.
These critics of pragmatics sometimes suggest that pragmatics in an argu­
ment is an empirical matter that is separate to each discipline. By these
lights, pragmatics is not a single field of inquiry in the evaluation of argu­
ments.
Still others argue that pragmatics is a kind of outer trimming around
the semantic core of argument. By these lights, classical logic is trimmed
with pragmatic maxims of conversational appropriateness. Yet another
alternative might be to argue that pragmatics represents something that
comes from the inside of semantics. By these lights, some principles of clas­
sical logic need to be reduced or altered in different pragmatic contexts. 9
Reasonable justifications have been given for each of these views. But
I propose here an alternative viewpoint that I think is more justified by
recent work in the field of argumentation studies. According to this view,
pragmatics already presupposes the existence of semantics, and would be
useless without it. Pragmatics is built around semantics and is an extension
of it. The basic concepts of semantics are truth and falsity. By classical
logic, these concepts are extended to include validity and consistency. The
basic concepts of pragmatics are assertion, retraction, and questioning.
These are defined in the context of a game of dialogue. A game of dialogue
is a set of players, moves, and procedural rules. These rules comprise locu­
tion rules, dialogue rules, logical rules, commitment rules and win-loss
rules. The logical rules define what constitutes a valid argument and a con­
sistent set of propositions. Hence semantics is included in pragmatics. The
remaining kinds of rules all function together to define assertion, retrac­
tion, and questioning for the different contexts of dialogue. An assertion of
a proposition ρ by a player a is made where, according to these rules, ρ is
added to a's commitment-set when it was not included there before. A
retraction of a proposition ρ by a player a is made where, according to the
rules, ρ is deleted from a's commitment-set. A yes-no question 'p?' is asked
by a where, according to the rules, where the other player b at the next
move either asserts p, retracts p, asserts lp, or makes an illegal move. A
why-question 'Why p?' is asked by a where, according to the rules, the
other player b at the next move either, (i) retracts p, (ii) asserts some prop­
osition q that implies p, where q is a commitment of a, or (iii) b makes an
illegal move. Then questioning is defined as asking a yes-no question or a
 THE LOGIC OF PROPOSITIONS 91

why-question. In some games of dialogue, other kinds of questions, e.g.

whether-questions, may be allowed as well.10
We have already studied subject-matter relatedness of propositions as
one kind of relevance. We could call it semantic relevance, although there
may be other kinds of semantic relevance as well. Semantic relevance is a
kind of local relevance — relevance between pairs of propositions.
By contrast, pragmatic relevance has to do with procedural rules in
games of dialogue. Pragmatic relevance is a kind of global relevance
because a whole network of moves in a sequence of dialogue may need to
be taken account of in order to determine the pragmatic relevance of one
particular move in the game. One kind of pragmatic relevance is the notion
of a relevant answer to a question in the context of a controversy or issue.
We will study this type of relevance in the fifth section of the next chapter.
The problem here has to do with how direct an answer must be to avoid a
reasonable criticism that it is "irrelevant."
Another type of pragmatic relevance has to do with the conclusion set
to be proved by a participant in dialogue. The fallacy here is one of getting
the wrong conclusion even if the argument may be valid. This was, of
course, one of the major faults of arguments we studied in chapter one. We
return to the analysis of this type of problem in the sixth section of the next
chapter.
Yet other types of criticism of irrelevance have to do with the charge
that there are too many premisses or not enough premisses. We have
acknowledged these criticisms as well, and will return to them.
The dialectical framework sketched out above, at least so I think, rep­
resents the most mature account of the relationship between pragmatics
and semantics. According to that account, semantics is included in pragma­
tics as its basis. Semantics is what makes a game of dialogue "logical." Prag­
matics is what makes a game of dialogue applicable to realistic contexts of
questioning and disputation of a thesis at issue. However, there is one
exception I must now make to this tidy arrangement.
In realistic arguments, the rules of dialogue are too often not defined
precisely. Our case studies of argumentation show that, in realistic contexts
like political debates, it is sometimes not clear what the issue is supposed to
be, or how the objectives of the dialogue are to be defined. But before an
argument can be evaluated for "fallacies," or for fairness of arguments, the
participants must have come to agreement on certain procedural matters at
the outset. What are the topics of the argument, what is the thesis each dis-
92 INFORMAL FALLACIES

putant is supposed to prove, and so forth. In many a realistic case of argu­

ment, however, these matters become themselves subject to dispute! In
such cases, there can be some latitude in choosing one set of rules over
another. Yet nothing can be constructively evaluated as a reasonable out­
come of the dispute until some clear and precise procedural rules have been
agreed upon by the participants in the argument.
Hence it is that a certain degree of conventionalism is tolerable in
reasonable games of dialogue. Some rules are more favorable than others
in certain contexts, but it is paramount that the arguers be brought to
accept some clear set of procedural rules if their dispute is to be reasonably
resolved or evaluated.
One implication of this tolerance of degrees of conventionalism in the
pragmatic framework is that in some contexts, non-classical logics may
reasonably be adopted as providing alternative logical rules for dialogue. In
some contexts, where relevance of subject-matter is highly contested, adop­
tion of relatedness logic as the core logic in a game of dialogue — instead of
classical logic — can not only be tolerated but positively approved. But
what does this mean in terms of the pragmatics versus semantics distinc­
tion? Have we cut down the semantic component to a subset of the classical
rules and added 'subject-matter overlap' to the pragmatics? Or would it be
more accurate to say that we have replaced the classical semantics by a non-
classical semantics? There is legitimate room for dispute here.
I am inclined to think that what we have done here is to keep part of
the classical semantics in place while adding the pragmatic dimension of
subject-matter overlap to our pragmatic structure. But I think that Dick
Epstein might legitimately disagree here, adopting the point of view that
relatedness logics, and their nonclassical cohorts, are really alternative
semantical models to classical logic. Not wanting to be dogmatic, I would
like to leave this issue, to some extent, open to further discussion. Related­
ness logic is fairly new on the scene, and much remains to be discussed
about its place in the scheme of things. I see it more as an applied logic —
a pragmatic structure — than its other proponents appear to. Much the
same disputed issue carries over to relevance logics generally. I see them as
applied logics appropriate to various contexts of criticisms of irrelevance in
the context of dialogue. However, I think the creators of these logics have
seen them in a very different light. They have tended to see them as alter­
native accounts of semantics to classical logic. They see classical logic as
incorrect (fallacious) and relevance logic as the "true" account of deductive
 THE LOGIC OF PROPOSITIONS 93

validity in all contexts of arguments. I tend to see classical logic as applica­

ble to some contexts of realistic argumentation — where relevance is not at
issue — and relevance logics as applicable to various other contexts of argu­
ment where criticisms of relevance are at issue.
It seems that it must remain the borders between semantics and prag­
matics be subject to further controversies. Nonetheless, I have striven to
make my position in the controversy as clear as it can be for the present.

10. What is a Fallacy?

We have now come to a point that needs resolution in our exploration
of the logic of the fallacies. It seems that we have found that formal logic
has its uses in helping us to construct criticisms and refutations of an oppo­
nent's position in argument, even though it also has clear limitations for
such purposes. Formal logic has to do with the internal truth-theoretic
structure of validity in arguments. True, formal logic can be extended to
encompass matters of relevance like subject-matter overlap of propositions.
But the fallacies we looked at in the first chapter — ad hominem, ad ver-
ecundiam and so forth — involved much broader notions of relevant defen-
sibility of positions than these narrow truth-theoretic or subject-matter con­
ceptions of validity could possibly be brought to bear on.
For example, the study of these fallacies in our sample parliamentary
debate involved alleged fallacies of question-asking. The issue there was
whether a certain reply to a question posed in parliament was "relevant" or
"unduly evasive." Also involved were questions of whether appeals to
authority were appropriate or relevant.
One of the problems in the development of applied logic is to get a
workable and reasonable notion of what a fallacy is. A fallacy is of course
an error in argument, or an incorrect argument, it is true. But virtually all
of the fallacies we have examined and will examine turn out to have correct
forms of argument as well. If so, there is nothing inherently or deeply nega­
tive about the study of the fallacies, even if the initial impetus to study a
particular type of argumentation is a concern with pathological instances of
it. Thus the goal of the study of fallacies and related specimens of interest­
ing argumentation should be the clear and reasonable sorting out of cases
into valid or invalid classifications, and not an exclusive preoccupation with
incorrect arguments per se.
Another problem is the psychologism inherent in the doctrine that a
fallacy is an argument that seems valid but is in reality not correct. Does
94 INFORMAL FALLACIES

this mean that in order to be fallacious, an argument must truly seem valid
to some reference group — a typical audience, most of the people some of
the time, at least one person — how do you choose? If so, the study of fal­
lacies is at least partly then a branch of psychology, or some social science.
In order to deal with this difficulty we need to distinguish clearly
between the correctness/incorrectness of an argument and its effectiveness/
ineffectiveness in persuading a target audience or disputant. The latter is
surely psychological, whereas the former is a normative question and surely
also to some extent a question of formal models or at least clearly laid out
decision procedures or rules. The confusion between these two questions is
often, quite rightly, inveighed against by formal logicians, who see their dis­
cipline as not psychological but mathematical.
I think the best approach to the difficulty is this. Surely a particular fal­
lacy does come to be of interest in studying applied logic precisely because
it is a tricky deception of argument that is important to study because it is
deeply and widely influential in commonplace instances of persuasion. The
more a particular type of argument, like appeal to arguments from exper­
tise, is widely influential in modifying the beliefs of actual participants in
persuasion, the more its claim to prominence as one of the "famous fal­
lacies" is assured or enhanced.
Thus the study of fallacies is indirectly linked to linguistics, psychology,
rhetoric, and other empirical sciences. But that doesn't mean that a particu­
lar instance of argument must seem valid to some particular person or audi­
ence to qualify as a fallacy. What is most important is that to be justified in
claiming that an argument is fallacious, or a fallacy, it is not enough to show
that the argument seems invalid to some actual persons. It is a normative
claim that the argument is in some sense bad, unsound, invalid, or at least
open to reasonable criticism. And that is a claim that must be justified by
an appeal to the relevant theory of (good) argument.
The demand of some commentators then that "actual instances" of fal­
lacies are mandatory and that artificial examples of any sort are inadmissi­
ble — even if they are moderately realistic, or strongly suggestive of
interesting steps of reasoning — is a confused and needless requirement.
The fact that Mary Brown enunciated a particular argument on May 18,
1982, on the Maryland Bridge at 8:30 p.m. does not suddenly elevate the
argument from mere theoretical abstraction of no interest to a "real fal­
lacy." The effectiveness aspect of fallacies is what makes the applied logic
of the fallacies "applied," but does not mean that psychological considera-
 THE LOGIC OF PROPOSITIONS 95

tions play a decisive role in the logical considerations of when an argument

is correct, incorrect, or even a distinct type of argument to be considered.
This however brings us to a main controversy in the study of applied
logic today. The question is to what extent formal logic is applicable to nat­
ural language conversational disputation. This question is fundamental to
our enterprise, and we will have to return to it again and again in this book.
The answer we will propose is that formal logic can be used by a partic­
ipant in disputations to build up his own argument or to criticize his oppo­
nent's argument. But in speaking of criticism in disputation we are import­
ing a framework, a conception of argument that includes more than just the
semantic structure of the propositions that make up the core of the argu­
ment. It includes as well the pragmatic structure of certain conventions or
rules of argument — locution rules, dialogue-rules, commitment-rules, and
strategic rules. The locution-rules define the types of locutions permitted as
moves in the dialogue. For example, certain types of questions, statements,
or withdrawals may be allowed. The dialogue-rules define admissible
sequences of moves. For example, a dialogue-rule may require that when
one player asks a question of the "yes-no" form, the other player must
respond at his next move by replying 'Yes,' 'No' or 'No commitment.' The
commitment-rules define how commitments are inserted into or deleted
from a store of commitments assigned to each participant in the dialogue.
The strategic rules define which sequences count as a "win" or "loss" of the
game.
This answer will be fleshed out more as we proceed. But two aspects of
it are of immediate note. First, the rules of the best sort of games incorpo­
rate a formal logic. It is the "language" of the players, and defines which
propositions follow by deduction from other propositions. Of course, one
can always ask which formal logic is the best for the logic of dialogue. But
we have already shown that classical logic, and its extension, relatedness
logic, can be useful in managing criticisms and refutations.
Second, psychologism is avoided by defining commitment-stores after
the fashion of Hamblin (1971) as sets of propositions. By this approach, you
can think of a commitment-set as a set of propositions written on a slate and
set out in public view of the participants and onlookers of the dialogue.
These propositions then do not necessarily represent the psychological
beliefs of the players.
By the above concept of argument, a fallacy is a type of move in a
game of dialogue that violates a certain rule of the game. Such a fallacy may
96 INFORMAL FALLACIES

be one of the kinds traditionally called an "informal" fallacy. Formal fal­

lacies are those that pertain to the formal logic element, the core of the
game that has to do with relations of validity in the set of propositions
advanced or withdrawn by the players. Informal fallacies have to do with
rules and procedures of reasonable dialogue.
Much turns then on what the best conventions or rules of dialogue are
in relation to the practical contexts of argumentation like the parliamentary
debate. The study of the applicability of these rules of dialogue, and their
formulation within the theory of logical dialogue itself, is the best means
towards the analysis of the fallacies.
The theory of the fallacies is not, at any rate exclusively, to be found in
formal logic. It is best pursued in the theory of argument, otherwise known
as logical pragmatics or the theory of logical dialogue-games.

NOTES

1. A discussion of logic as a device for creating new arguments — the logic of discovery — is
included in the last section of chapter eleven.
2. An arguer's set of commitments is said to be ambivalent if he becomes committed to a
proposition ρ during the course of an argument, but then, during the same argument, indicates
that he is not committed to p.
3. Massey (1981) makes use of this fact to cast doubt upon the use of formal logic to study
informal fallacies. Massey's arguments are criticized in Walton (1985, ch.l).
4. An account of the notion of subject-matter relatedness is given in Walton (1979) and a for­
mal analysis of relatedness logic is given by Epstein (1978).
5. A much fuller account of paradoxical forms of inference in classical and relatedness prep­
ositional calculi is given in Walton (1982).
6. Other possibilities are allowed by Epstein (1979). However, the simple approach given
here is very natural, and is adequate for our present purposes.
7. These sorts of restrictions have already been discussed in chapter one, but they are elabo­
rated on in subsequent chapters.
8. As we saw in chapter one, an argument can be criticized because it has too many premis­
ses, too few premisses, the wrong conclusion, and so forth.
9. In connection to the question of relevance, we might distinguish between local relevance in
an argument, e.g. subject-matter overlap of the propositions, and global relevance. Global rele­
vance pertains to pragmatic questions of "where the argument is going" as the unfolding of a
controversy, issue, or dispute.
10. The next chapter will give detailed examples of formal games of dialogue with question-
answer rules.
 CHAPTER 4: LOGICAL DIALOGUE-GAMES

The conversational quarrel and the forensic debate are dynamic and
interpersonal processes that capture in a lively way the cut and thrust of
realistic argumentation. The approach of the formal logic of propositional
inference is, by contrast, an impersonal, static conception of argument, one
that essentially views an argument as a set of propositions. Granted that a
set of propositions constitutes the core of an argument, can we not follow
up the suggestion of these more lively processes that there could be more to
an argument than just the set of propositions that makes up its essential
core? The problem with the quarrel and the debate however, is their essen­
tial subjectivity. Being adversary procedures, and only partially regulated,
disputes within them cannot be clearly resolved on the point of whether a
given argument is to be adjudged correct or fallacious.
By contrast, in deductive logic we have a clear methodology, an objec­
tive way of determining whether the argument once given is correct or
incorrect by set standards. In order to gain some of the advantages of all
these models of argument and at the same time overcome the essential
restrictions of each, in this chapter we take the approach of viewing an
argument as a logical dialogue.
Basically, a logical dialogue consists not only of a set of propositions
but also a set of questions that stand in certain relationships to these propo­
sitions. Moreover each question and its related proposition (answer) is
indexed to a participant in the dialogue who may be at different times an
asker or an answerer of a question.
Formally speaking, the dialogue is a game. That is, a dialogue is
defined by a set of rules which regulate permissible moves that the partici­
pants can make, and each move consists of a question or a proposition.
Thus the core of a game of dialogue is a set of propositions and questions,
each of which is indexed to a participant.
For each game of dialectic1, rules are defined that determine what con­
stitutes a set of winning moves and a set of losing moves for each partici­
pant. Dialectic can be studied in a purely formal manner, as in the sort of
98 INFORMAL FALLACIES

game two computers might play, where rules and moves are all precisely
defined. But it is also possible to have realistic games of dialectic that are
only partially formalized, and where the sequence of argumentation corre­
sponds more nearly to what most of us would consider natural argumenta­
tion. Outright naturalism might take place in a quarrel or debate. Thus,
many an interesting yet fairly well regulated sequence of dialectic would
occupy a position halfway between the purely formal game of dialectic and
the very realistic but very loosely regulated debate or quarrel. Thus the
dialectic approach of utilizing formal games of dialogue shows some prom­
ise of effecting a compromise between realistic "actual" conversational
argumentation and abstract formal models of valid or invalid argument.

1. Different Approaches to Formal Dialogues

The original conception of the dialectical model of argument is due to
Aristotle, who formulated it roughly as follows. First there is to be one
questioner and one answerer in a game of dialogue. Second, the answerer
is obliged to defend some thesis on a topic. Third, the questioner puts for­
ward probing critical questions. Fourth, the answerer is then obliged to
answer as clearly and straightforwardly as he can, these questions, one by
one. It is required of the questioner that he or she ask clear and straightfor­
ward questions. If a question is misleading or unclear, the answerer must
have the option of not responding and perhaps also the option of asking
that the question be clarified. The answerer must try to give as direct
answers as he can to the question and not attempt to use specious means to
avoid answering the question. However, if he truly cannot answer the ques­
tion through ignorance, he should reply that he does not know the answer
and the questioner must go on to formulate new questions that may be able
to overcome that ignorance. If the answerer provides an answer that con­
tradicts one of his previous answers, then the contradiction should be dealt
with, perhaps by removing one of the answers to eliminate the contradic­
tion.
According to Aristotle's conception, such a game of dialogue con­
cludes if the answerer is refuted or if it is clear to both participants that the
questioner's refutation will not succeed. This outline of Aristotle's basic
conception of dialogue gives a general indication of the shape, character
and intentions of such dialectical games. But as the framework is stated
above, there remains considerable room for negotiating how precise rules
should be formulated in particular types of games. We will see subsequently
 LOGICAL DIALOGUE-GAMES 99

that there are particular types of games which have quite different proper­
ties, but all of these games will share the general characteristics of dialectic
outlined above by Aristotle's model.
According to Hamblin (1970, Chapter 8) an important aspect of games
of dialectic concerns the nature of the commitments made by the partici­
pants in their asking and answering of questions. Hamblin constructs a type
of game of dialectic which we will here call the game H which has two par­
ticipants, White and Black. The language of the statements advanced at
each move is that of classical propositional calculus. Each participant in the
game can ask or answer questions, and in so doing they commit themselves
to certain propositions by means of the way the rules for asking and answer­
ing questions are stated. The answering of a question by means of advanc­
ing a certain proposition ρ automatically places ρ in the answerer's commit­
ment-store, according to the rules of H. The asking of a question 'p?'
automatically puts ρ in the asker's commitment-store, but only puts it in the
answerer's commitment-store if he fails to indicate that he is not committed
to p. In this way then, both participants in the game build up a stock of
statements in their respective commitment-stores.
One of the most useful resources for the study of fallacies in dialogues
is the logic of questions, the formal theory of questions and answers devel­
oped in recent years by Åqvist, Belnap, and Hintikka. The pragmatic appli­
cations of the theory of questions to the context of interactions between a
questioner and an answerer in acts of communication have been developed
by David Harrah (1963), (1980) and (1984). Harrah has developed formal
models of communication where a sender's messages arrive over a period of
time. The receiver can reject some messages, accept others, change his
mind in various ways, and can apply various evaluation functions to build
up a "usable message total." Messages are evaluated in light of the
receiver's background knowledge.
Harrah's proposal, that getting a usable message in communication
involves selecting out maximal consistent subsets on the basis of the order
that the messages are received, is an interesting one in relation to the prag­
matics of dialogues. In conversational arguments how would the order of
the communication events affect the management of inconsistent sequences
of messages? There would seem to be several possibilities.
On some occasions, the later message might be the more acceptable,
where it conflicts with a previous message. The reason might be expressed
as follows: "Since you now say such-and-such, given that it conflicts with
100 INFORMAL FALLACIES

what you said earlier, such-and-so, this must mean that you have decided to
give up your previous viewpoint." In another context, quite the opposite
ordering could be the more plausible: "What you are saying now cannot be
accepted, because it conflicts with the basic position you started with at the
beginning." By these lights, the earlier commitment could be, in some
sense, the more fundamental in the development of argument.
The problem is the classic one for the theory of dialogue. When con­
fronted by an inconsistency in an arguer's collective moves, how should
retraction be most reasonably organized where (1) retraction is allowed in
the game generally and (2) where retraction is the appropriate move for a
player who has sent out a set of propositions that turns out, or is shown by
the other player, to be inconsistent.
A formal theory of questions and assertions in dialogue developed by
Ruth Manor (1979), (1981), (1982), and (1983) studies pragmatic properties
of an utterance as a function of the context of dialogue. Assertions are con­
sidered as pairs of questions and answers. The context of an assertion is
determined by the question "under discussion." Thus notions of relevance
and commitments of answers are determined by the moves a player makes
in the question-answer game. Manor (1983, p.71) has commitment sets for
participants in a game of dialogue, and uses this device to study problems of
retraction and assertion.
According to Hamblin a commitment is not necessarily a belief of the
participant in argument. Rather, Hamblin thinks of a commitment-store as
a set of tokens of statement or propositions that the player becomes com­
mitted to through the way he plays the game. Thus, according to Hamblin,
a good way to conceive of a commitment-store would be as a sheet of paper
with a number of sentences written on it, or perhaps as a memory bank of
a computer.
As the dialogue proceeds, propositions are then added to this commit­
ment-store, or in some cases if rules allow for retraction, propositions can
also be deleted from the commitment-store. It is important to note that in
a Hamblin game, if a participant is committed to a certain proposition he is
not thereby always automatically committed to all the logical consequences
of that commitment. Moreover, according to Hamblin, commitment-stores
need not always be consistent. As Hamblin sees it, these requirements of
implication and consistency can be varied with the type of game that is
involved. Thus as Hamblin sees games of dialectic there is considerable
room for flexibility in how the rules of a game can be set up and how the
 LOGICAL DIALOGUE-GAMES 101

commitment-store operations can be organized.

According to Hintikka (1979) a symmetrical model of dialogue can be
given that incorporates rules for questions and anwers. In a Hintikka game,
there are two participants, α and β, and each puts forward an initial thesis
A0 and B0 respectively. In order to prove his thesis, each player is allowed
to use only two kinds of premisses. First each player may use the other
party's initial thesis. Second, each player may use the answers given by the
other party to one's own questions. Using these premisses, the object of
each participant is then to win the game by proving his thesis according to
the rules of logic allowed by the game.
Hintikka allows that some particular games of dialogue are not of this
symmetrical type. In some games, one party is required to answer questions
which the other party is restricted to asking. This type of game, familiar
from the example of the Socratic dialogues portrayed by Plato, is a one­
sided or a symmetrical type of game. In this type of game, usually called the
obligation game, the object of the questioner is to try to trap the answerer
into inconsistent concessions. He wins if he succeeds in doing this. How­
ever, his opponent wins if he fails to achieve this object, after an agreed
upon finite number of moves.
According to Hintikka, the symmetrical model of dialogue is especially
favorable as a foundational beginning for the study of dialectic because
other types of games can be shown to be special subsystems of the basic
symmetrical game. Sometimes in a Hintikka game, the objectives of the
players are incompatible in the sense that the thesis of one is the negation
of the thesis of the other. In this type of game, to prove one's own thesis is
in effect to refute the opponent's thesis. Therefore, in this basic type of
game the two players are directly opposed. Hintikka calls this special type
of game a dispute.
In short, the Hamblin games and Hintikka games are basically similar
in general outline, and both come under the general umbrella of Aristotle's
conception of dialectic. There are differences of detail between the two
approaches. But the main difference concerns the objective of each type of
game. According to Hamblin (1970, p. 137), "the purpose of the dialogue is
the exchange of information among the participants." What precisely con­
stitutes "exchange of information" is not sharply defined by Hamblin. This
means that Hamblin games have a certain open-ended quality to them.
However, Hamblin's assumption that dialogues should be "information-
oriented" (his term, p.137) has definite implications for the design of the
102 INFORMAL FALLACIES

rules of his dialogue-games.

The basic feature of the Hamblin game (H) of special interest here is
that an answerer has a no-commitment option. If asked a yes-no question,
a player may reply "No commitment' as an allowed alternative to 'yes' or
'no.' If asked 'Why p?', a player may reply 'No commitment p' as an
allowed alternative to offering justification for p. This openendedness is
quite a good thing when it comes to some of the fallacies. For we have
already seen in several of our case studies that an answerer should not
always be forced into a position of having to give a direct answer to every
question in a reasonable dialogue.
The Hintikka conception of dialogue is much less open than the
Hamblin game (H). In the Hintikka game, the 'No commitment' response
is not fully available to an answerer — indeed, the answerer may be forced
by the questioner to commit himself either to the presupposition of a ques­
tion or the negation of the presupposition. This convention certainly speeds
the game along, but it gives the questioner a good deal of power, and
thereby makes the answerer relatively powerless in the game.
In a Hintikka game, an answerer may refuse to answer a question. But
if he refuses, the negation of the presupposition of the question is
immediately added to his commitment-store. For example, if an answerer
refuses to answer the question 'Who lives in that house?,' he immediately
becomes committed to the proposition 'Nobody lives in that house.' Or if
he refuses to answer the question, 'Is this zebra black or white?,' he
becomes committed to the proposition, 'This zebra is neither black nor
white.' So the questioner can press forward with aggressive questioning in
this sort of game and get some revealing answers.
The other significant feature of the Hintikka game is that win-loss
objectives for the players are precisely defined. The player wins who first
deduces his thesis, by means of the logical rules, from the other player's
concessions. This means that the Hintikka games are not open-ended in
their ultimate objective in the way that the Hamblin games are.
A third approach to logical games of dialogue is that of the Erlangen
School, based on the insights of Paul Lorenzen, and including the collab­
orations of his colleagues and students Kuno Lorenz, Wilhelm Kamlah, and
Oswald Schwemmer. The rules for Lorenzen dialogues are even more
tightly defined than for Hintikka dialogues. This tightness is achieved by
making the dialogue-rules isomorphic with rules for a formal deductive
logic. In effect, the rules for deductive logic, e.g. propositional calculus, are
 LOGICAL DIALOGUE-GAMES 103

re-interpreted as dialogue rules for the players' moves in a game.

Kamlah and Lorenzen (1967) presume that in a game of dialogue, the
participants have access to a method of verifying simple (elementary) sen­
tences. This checking is carried out by expert and rational participants in
the game who are able to carry out the objective process of verification in a
correct manner. The process is called "interpersonal verification," and it
means that in a Lorenzen game of dialogue, the truth-value of the simple
sentences can, at least in principle, be established in the game. However,
the point of the game is for the players to determine how the complex sen­
tences are built up from these simple sentences by means of rules for the
logical constants.
Each of the logical constants for propositional calculus is defined
dialogically as follows in Lorenzen (1969). If a player (the proponent)
asserts the conjunction   , then the other player (the opponent) wins if
he successfully attacks either A or . The proponent wins only if he suc­
cessfully defends the attacks on both A and B. If the proponent asserts the
disjunction A v B, then the opponent must mount two successful attacks in
order to win. He must attack both A and B. However, if the proponent suc­
cessfully defends either A or B, then he wins.
Now suppose the proponent asserts a conditional, A → B. When the
opponent attacks this conditional, he is obliged to assert A. If he fails to
successfully defend A against the proponent's attack, he has lost. If he suc­
ceeds in defending A however, then he must go on to defend B. If he suc­
ceeds in defending , he wins. Otherwise — if his defence fails — he loses.
Finally, there is the rule for negation. Suppose the proponent asserts
1A. Then if the opponent successfully defends A, he wins. If the opponent
is unable to defend A successfully, then the proponent wins.
Given the above four dialogical definitions of the logical constants, two
kinds of dialogue-games are outlined by Lorenzen. In a material dialogue-
game, winning depends on successfully defending an elementary (simple)
sentence, in accord with the following three dialogue rules called conven­
tions by Kamlah and Lorenzen. First, the proponent starts by asserting a
thesis. Second, the proponent wins if he successfully defends an attacked
elementary sentence, or if the opponent fails to successfully defend an
attacked elementary statement. Third, at each move, each player must
either attack a previous move of the other player, or defend against a previ­
ous move of the other player. In this type of material dialogue-game, the
assertion of complex sentences ultimately decomposes into the assertion of
104 INFORMAL FALLACIES

simple statements, according to the rules. Hence the ultimate question of

who wins depends on the successful attack or defence of a simple sentence.
By contrast, in a formal dialogue-game, the question of who wins does
not depend on the simple sentences. In these games, the compound sen­
tences asserted or attacked by the players are tautologies or inconsistencies.
This sort of game involves modifications of the three rules of dialogue given
above for material dialogue-games. Kamlah and Lorenzen regard the for­
mal dialogue as an abstraction or formalization of the material dialogue.
The Lorenzen games are closer to Hintikka games than to Hamblin
games in the respect that there are precisely formulated win-loss rules in the
Lorenzen and Hintikka games. Yet in another way, the Lorenzen games
are quite different from either of the other two types of dialogue-games. In
the Lorenzen games, every move begins with an assertion, then continues
in a regulated sequence of attacks and defences. Once the initial move is
made, the whole remainder of moves in the game is regulated so that nei­
ther player is allowed to make any new assertions — at penalty of losing —
until the game is won or lost. By contrast, the Hamblin and Hintikka games
allow for the asking of new questions, or the introduction of new state­
ments, at various points in the game.
Moreover, the Hamblin and Hintikka games both involve sequences of
question-asking and question-answering moves. The Lorenzen games are
much simpler and less open in the sense that they do not really involve the
asking of questions by the participants at all. In a Lorenzen dialogue, there
are only assertions, and then certain responses to those assertions, as deter­
mined by the rules for complex sentences.
We could sum up the main differences in the three approaches as fol­
lows. In the Hamblin and Hintikka games, there is a core propositional
logic, but surrounding this core, the outer dialogue-rules permit considera­
ble freedom to the players in asking and anwering questions. However,
there is less freedom in the Hintikka game, because win-loss rules are
exactly formulated. In the Lorenzen games, the whole game is taken up by
the propositional logic — that is, by the rules for the complex sentences —
so there is no outer structure of the game beyond the three types of
dialogue rules. These rules permit little freedom of movement to the
players. In the formal dialogue-games, win or loss is exclusively determined
by the rules for the logical constants. In the material dialogue-games, win
or loss is determined by the truth or falsity of the simple sentences as estab­
lished by the "rational experts," and by the rules for the logical constants.
 LOGICAL DIALOGUE-GAMES 105

2. The Ad Ignorantiam Fallacy

We have seen already that a number of the fallacies relate to the asking
and answering of questions. Consequently, one of the first proving grounds
for any theory of dialogue, in relation to the fallacies, must be its adequacy
in the reasonable management of the process of question asking and
answering. Therefore, a good place to start in the theory of dialogue will be
for us to ask how each of our three major approaches to dialogue-games
regulates the process of questioning.
Let us examine each of the three approaches individually, then turn to
comparisons. We start with the Hamblin games.
Without setting up all the elaborations needed for a fully regulated for­
mal dialogue game, 3 it will be enough for our purposes to adopt some ele­
ments of a Hamblin (1970) game in order to illustrate some questionable
sequences of moves. In this type of game, called by Hamblin a "Why-
Because-Game-with-Questions," each player may ask questions of the form
'Why p?' The rule for answering why-questions in the game requires that
the answerer either (a) furnish another proposition that implies ρ by modus
ponens, e.g. 'q ⊃ ρ, q, therefore ρ,' or (b) indicate "no commitment" to p.
Such a rule immediately disallows the following type of ploy.

This sort of move in a dialogue-game is quite generally interesting, and in

fact models a traditional type of fallacy called the ad ignorantiam fallacy.
This fallacy is said to occur where an arguer tries to prove some proposition
ρ by arguing from the premiss that ρ has never been disproven. For exam­
ple, someone might try to argue that extra-sensory perception must exist
because nobody has ever been able to prove that it doesn't exist. In another
form, one argues ad ignorantiam by attempting to conclude that ρ is dis-
proven because ρ is not proved. For example, someone might argue that
extra-sensory perception does not exist because nobody has ever been able
to prove definitively that it does exist.4
According to the dialogical model of this fallacy represented above,
the error has to do with a shift in the burden of proof. When White asks
"Why p?", the burden of proof is put on Black to furnish proof for ρ in the
form of some proposition that implies p. Black, however, tries to shift the
burden of proof off herself and back onto White by responding with a ques-
106 INFORMAL FALLACIES

tion "Why ]p?" Clearly the problem with allowing this sort of move in
games of dialogue is that it would make it too easy for an answerer to
always avoid answering a question.
This points up a general problem about Hamblin games. What do you
do with the "skeptic" who, in answer to any question always replies "No
commitment." The Hamblin game (H) allows an answerer the freedom of a
"No commitment" reply to any question and that, as we saw, could be a
good thing. For if the answerer were not allowed some option of this sort,
a form of ad ignorantiam fallacy could be built into the game. Suppose a
questioner asks 'p?' Without the no-commitment option, an answerer
would have to reply 'p' or 'lp'. But suppose in fact he doesn't have good
evidence for knowing whether ρ is true or not. If forced to commit himself
immediately to ρ or to lp, he is in effect committing a form of the ad
ignorantiam fallacy. Surely the rules of the game should not be designed to
force an answerer to commit a fallacy. Consequently, the availability of the
no-commitment option in (H) is generally a good thing as a feature of
reasonable dialogue. But a problem results. How do we contend with the
answerer who filibusters by always replying 'No commitment' in answer to
any question he wishes to avoid?
In a Lorenzen dialogue, this problem can never arise at all, for there is
no freedom of asking questions or answering them, at least in the same way
allowed in a Hamblin dialogue. Once an assertion is made, it must be
defended or the proponent loses. The opponent must attack it, or he loses.
The problem of avoiding commitment can never arise in the sequence of
moves in a Lorenzen game. However, the problem is avoided at some cost.
According to the negation rule in a Lorenzen game, when the propo­
nent asserts a proposition ]p, the opponent can win only by asserting, and
then successfully defending p. This means that if the proponent asserts
"There are no ghosts," the opponent must assert "There are ghosts" and
then prove it. But suppose, in fact, that the opponent has no evidence
either to prove or refute the existence of ghosts. Nevertheless, in a Loren­
zen game, he must be committed to the assertion "There are ghosts" and
prove this assertion. In effect then, he must commit a form of the ad
ignorantiam fallacy. Of course, the Lorenzen game presupposes that "ra­
tional experts" provide evidence for at least the simple propositions in the
game. But does that assumption help us here? I do not see how it does.
This type of problem is indicated by the quantifier rule given by Loren-
 LOGICAL DIALOGUE-GAMES 107

zen (1969, p.24): if the proponent asserts l(Vx)Fx, then, the opponent may
assert (Vx)Fx and the proponent will ask him to defend it. If the defence
succeeds, the proponent loses. Otherwise, the proponent wins. Lorenzen
adds that if the opponent does not know how to defend (Vx)Fx, it is advis­
able for him not to challenge the original thesis l(Vx)Fx. This evidently
means that in the Lorenzen game of dialogue, if the opponent does not
know whether (Vx)Fx is true or not he can't make any move in the game.
If he challenges l(Vx)Fx at all, he must prove its negation (Vx)Fx or lose
the game.
This solution is the diametrical extreme of Hamblin's. Hamblin's
answerer had virtually complete freedom. Lorenzen's challenger really has
no freedom. If he can't refute, he'd better not even challenge an assertion
or raise the question of whether it might be true. I think neither of these
ways of managing ad ignorantiam sequences of dialogue is adequate to the
study of ad ignorantiam as a fallacy or criticism of arguments.
Hintikka's concept of dialogue represents a midpoint between the two
extremes of absolute freedom and absolute constraint. According to Hin­
tikka (1979, p.237), if the answerer refuses to answer a question, the pre­
supposition of the question is added to his commitment-store. But we will
see, and we have seen, that this approach can lead to problems of its own.
It means that the questioner has the power to make the answerer become
committed to either the presupposition or its negation of any question that
he (the questioner) may choose to ask.
When we come shortly to a discussion of the fallacy of many questions,
it will be clearly evident that this approach gives too much power to the
questioner. The problem is even evident with simple yes-no questions. The
simple yes-no question p? has as its presupposition the tautology ρ ν lp.
But the negation of that presupposition (in classical logic) is equivalent to
the contradiction ρ ^ lp. Therefore, in a Hintikka game with rules strong
enough to yield classical logic, any answerer who refuses to answer any yes-
no question immediately becomes committed to a contradiction, and
thereby loses the game.
In short, the Hintikka conception of dialogue is the best of three
approaches in managing ad ignorantiam argumentation, but it still has some
severe difficulties as solution by itself. The solution I will propose in this
chapter involves accepting Hintikka's general conception of logical
dialogue, but modifying it in a certain way. Before going to this new type of
108 INFORMAL FALLACIES

game based on the Hintikka conception however, we must look to some

other fallacies that have to do with question-answer dialogue.

3. Fallacies of Question-Asking
Now we have raised the topic of questions, it may be somewhat discon­
certing to realize that not all fallacies involve propositions. Sometimes the
mere asking of a question can be fallacious, or at least logically question­
able. For example, suppose the prosecuting attorney in a criminal trial asks
the defendant the question, "What did you use to wipe your fingerprints
from the gun?" This question has several presuppositions that the answerer
may not want to commit himself to. For example, it has the presupposition
that the defendant wiped his fingerprints from the gun. It has the presup­
position that the defendant did this wiping at a certain time. It therefore
also has the presupposition that the defendant handled the gun, perhaps
also the presupposition that the defendant removed these fingerprints for
some suspicious or criminal purpose. If the defendant is in fact innocent,
and never at any time handled or even saw the alleged weapon, he may be
hard pressed to know how to answer or at least try to avoid answering the
question. If so, the defendant will probably say something like, "I didn't
ever wipe my fingerprints from the gun with anything at any time." How­
ever, we should notice that the prosecuting attorney is directing the respon­
dent away from this sort of answer by phrasing the question in such a way
that the citing of some particular item (used to remove fingerprints) is
required to qualify as a direct answer to the question. If the defendant has
never even seen the gun in question, he may feel that the attorney is being
somewhat unfairly aggressive in his questioning.
When evaluating a real question-answer segment of dialogue for possi­
ble fallacies or criticisms, one needs to evaluate the answer. But first, one
must look to analyzing the question. What type of question is it? A simple
yes-no question has a safe presupposition. But many other types of ques­
tions may have significant presuppositions that should be carefully enumer­
ated at the outset. If a question has several presuppositions, it may be many
questions rolled into one. Other factors to be considered are included in the
schema below.
 LOGICAL DIALOGUE-GAMES 109

There is nothing wrong or fallacious, as such, with a question that has sev­
eral presuppositions. For example, "Is the man at the back of the room
wearing the red sweater and glasses looking at his neighbour's examination
paper?" may be a non-fallacious question in some circumstances. It has a
conjunction as a presupposition. But that does not, by itself, make the
question fallacious.
However, if a question has multiple presuppositions, it is worthwhile
being clear that several presuppositions are involved. It may be a good idea
to list them. Next, it is worth asking whether any of the presuppositions are
prejudiced or unfair. For example, "Are you going to resign your office, in
view of your scandalous and illicit conduct in recent weeks?" contains pre­
suppositions that many would-be answerers should not fairly be asked to
accept. Whether a presupposition is fair or non-prejudiced is a matter rela­
tive to the particular circumstances of the answer. The above question
could be fair if "scandalous and illicit conduct" has already been conceded
by the answerer. Otherwise, the question may be unfairly aggressive and
prejudicial.
The procedure for answering has to be handled with extreme care in
relation to questions that have multiple presuppositions. For example, sup­
pose the prosecuting attorney asks the defendant the question, "Did you
buy this gun and then use it to shoot the victim fatally?" Let's assume that
it's true that the defendant did buy the gun but did not use it to shoot any­
one. Then it is difficult for him to straightforwardly answer this question.
What he needs to do is divide the question in two parts and say 'yes' to the
question concerning whether he bought the gun and 'no' to the question
concerning his alleged use of it to shoot someone. In some instances, how­
ever, it is clear that the questioner has constructed the question in such a
no INFORMAL FALLACIES

way as to attempt to allow the answerer no choice in his separation of the

presuppositions and thereby wants to trap the answerer into conceding
something that may be dangerous to the answerer.
Indeed, there are even certain real situations where the asking of ques­
tions is rigidly formulated in such a way that the answerer has no choice but
to give a single answer to a multiple question. It is well known for example
that bills as "riders" are often tacked onto proposed legislative bills so that
the opposition has no choice between two bills, A and B. The proposal is
constructed in such a way that the opposition has to vote on propositions A
and  altogether. So that it only has the choice of either rejecting both A
and  or accepting both A and B. In other words, if the opposition party
wants to accept A, a piece of legislation which it likes, it also has to accept
B, a piece of legislation that it does not like. This nasty situation then, is
similar to that of the answerer of the multiple question. 5
This is a good time to note that avoiding giving a direct answer to a
question is not always unreasonable, much less "fallacious." For if the ques­
tion is highly prejudicial, the best answer may be to challenge its presuppos­
itions. One may do this by answering the question with another question.
Alternatively, one may simply refuse to answer the question at all. Both
these kinds of responses may be quite reasonable if the question is not a fair
one. Hence the answerer should consider whether an answer is appropriate
at all. If it is not, it may be better to criticize or challenge the question itself.
Moreover, one may simply not know the answer. Or one may be able
to give an informative response without giving a direct answer, perhaps
because certain information is unavailable, or should not be given. Such
refusals to give direct answers may, in many instances, be reasonable and in
no way fallacious, either as ad ignorantiam or ignoratio elenchi fallacies.

4. The Fallacy of Many Questions

The most notorious of this group of fallacies of question-asking is the
so-called fallacy of many questions or complex question fallacy. This tricki-
ness is embodied in the famous question, "Have you stopped beating your
spouse?" The presumed intent of asking this question is to speciously force
the person to whom it is addressed into the position of conceding an unwel­
come proposition, no matter which answer is given. In effect, the question
permits only two possible simple and direct answers, 'yes' or 'no'. But
either answer implies what is presumably unwelcome to the answerer,
namely the proposition that he or she has at some time beaten his or her
 LOGICAL DIALOGUE-GAMES 111

spouse. If you answer 'yes' to this question, then you have committed your­
self to the proposition, 'I have stopped beating my spouse,' which implies
that at some time or other you have beaten your spouse. However if you
answer 'no' to the question, you are accepting the proposition 'I have not
stopped beating my spouse,' which again implies, just like the 'yes' answer,
that at some time or other you have beaten your spouse. So the answerer is
trapped. No matter which alternative is chosen, yes or no, she is committed
to the same unwelcome implication. What may appear to be a genuine
choice of alternatives is in reality a trap.
One has to be careful in the analysis of this fallacy, however, to note
that it may be a reasonable presumption in many contexts — but it is still
only a presumption — that the answerer has not at some time or other
beaten his or her spouse. If in fact, the answerer has at some time engaged
in spouse-beating and does not mind admitting it, then for that answerer in
that situation there is really nothing objectionable or fallacious about the
question. The answerer can then say, "Yes, I have stopped" or "No I have
not stopped," and the implication of either of these responses is not, we
may presume, objectionable.
However, it is for the non-spouse-beater, or at least the person who
does not want to acknowledge past occasions of spouse-beating for whom
the question poses a difficulty, and against whom the questioner commits
sharp practice by the asking of this question. For the innocent non-spouse-
beater is trapped no matter which way she answers and cannot see any obvi­
ous way clear — at least without avoiding a direct answer to the question —
to avoiding the unwelcome concession of having engaged in spouse-beating.
Of course, where the answerer is not forced to give a direct answer,
that is a 'yes' or a 'no,' as in most commonplace argumentation, the way to
handle the question is obvious. One should simply say in response to the
question, "No, I have never beaten my spouse at all. I haven't stopped, nor
had I ever started". This way of replying is in effect a rebuttal of the presup­
position of the question.
So what is really fallacious about the question? 6 The thing to notice is
that the question is fallacious because it is designed in such a way as to call
for a 'yes' or a 'no' answer. That is, the answerer who rebuts the presuppos­
ition of the question as we just indicated, is failing to give a direct answer to
the question. Thus it is because the question is put in the 'yes-no' form,
thereby calling for a direct 'yes' or 'no' answer, that the question is objec­
tionably fallacious. Because it is a yes-no question, it appears to be inno-
112 INFORMAL FALLACIES

cent, whereas in fact it is not innocent at all, because if the answerer does
directly answer it with a 'yes' or a 'no,' she commits herself to some unwel­
come, and in this case, incriminating proposition.
According to Belnap (1963, p. 127), a question is called safe if its pre­
supposition is logically necessary. For example, the yes-no question 'Is
snow white?' has as its presupposition the proposition 'Either snow is white
or snow is not white.' This proposition has the property of being a tautology
in classical PC. That is, no matter what value the proposition "Snow is
white' has, whether it is true or false, the whole proposition, 'Snow is white
or snow is not white' has to be true. It is always true regardless of whether
its component proposition is true or false. Therefore the yes-no question 'Is
snow white or not?' is safe in Belnap's sense.
According to Belnap all yes-no questions have this property of being
safe. That is, the presupposition of a yes-no question is logically necessary.
It is a tautology and is therefore presumably harmless. However, the
spouse-beating question, although it has the superficial form of a yes-no
question, and therefore is meant to appear safe, is by no means safe at all.
In fact, it is a very risky question. It has a presupposition that is not only not
logically necessary, it is in fact quite unwelcome and even potentially
incriminating to the answerer. Hence we can see that what is really falla­
cious about the spouse-beating question is that because of its superficial
logical form as a yes-no question, it has the appearance of being safe. But
if we analyze it more deeply and look at its underlying structure, it is not a
safe question at all. Thus the fallaciousness of the spouse-beating question,
in at least one important respect, is that it masquerades as something that it
is not. It masquerades as safe when it is in fact, risky.
Another fallacious aspect of the spouse-beating question has to do with
its being a loaded question. We presume that the answerer to the spouse-
beating question does not want to assent to or commit herself to the propo­
sition that she has at some time or other engaged in spouse-beating. Hence
this proposition may be said to be unwelcome to the answerer in the sense
that, in a given context of a game of dialogue, we may presume that that
answerer would not commit herself to the presupposition. However, not
only does the spouse-beating question have an unwelcome presupposition,
but what we mean by saying that it is loaded is that no matter which way she
answers the question, the answerer must thereby become committed to that
presupposition. The trickiness of the question then is that it forces the
intended victim to accept the unwelcome presupposition, no matter which
 LOGICAL DIALOGUE-GAMES 113

direct answer is selected. It is like the wellknown frustrating questions of

objective examinations that require the student to give an answer to a given
set of alternatives but do not allow the option "none of the above." Not
only is the question loaded but all the chambers are loaded.

5. Demanding Direct Answers to Questions

A central problem in the design of formal games of dialogue is to have
procedures for the asking of questions that allow the answerer to avoid
these different fallacies and presumptions implicit in the asking of a ques­
tion. That is, fair procedures for the handling of questions and answers
must be designed in such a way that the aggressive questioner is not given
too much power to unfairly browbeat the answerer and thereby win the
game too easily. However, the answerer must nevertheless have to confront
questions directly enough so that she is not given so much latitude that she
can always indefinitely put off answering the question, and thereby too eas­
ily win the game by default.
When one begins the study of fallacies, there is an understandable ten­
dency to criticize any answer that is not a direct answer to a question as a
"fallacy of irrelevant answer." This tendency is perhaps understandable,
given the advice of textbooks that "irrelevance" is a fallacy. But one must
be careful to realize that with many questions, simply offering a direct
answer may be very dangerous and misleading, especially if the question is
complex or unfair.
One problem is that in realistic dialogues, it is not always possible to
define what should reasonably count as a direct answer according to some
set of precise rules of a formal game of dialogue. In the games of dialogue
we have studied here, questions are restricted to yes-no questions or why-
questions. In these two cases, the rules clearly specify exactly what counts
as a direct answer. For a yes-no question, a direct answer is 'yes' or 'no.'
For a why-question, a direct answer is to provide some set of premisses that
implies the proposition queried by the logical rules of the game. But take a
question like "Where is my red shirt?" posed to my wife. She might answer
"In the house" or "In the bedroom" or "In the dresser" or "In your dresser
by the bed" or "In the top drawer of your dresser by the bed" or "In the left
side of the top drawer of your dresser by the bed, under the socks". Which
answer should count as the direct answer? Clearly, it depends on the cir­
cumstances, knowledge and needs of the particular questioner and the
answerer in this particular situation. Hence, beyond the simple yes-no and
114 INFORMAL FALLACIES

why-questions of the formal games of dialogue, defining 'direct answer' is a

substantive problem for the logical pragmatics of question-answering.
The problem here is similar to that of the design of any other type of
game. One does not want to weight the rules of the game too heavily
towards one side or the other so that it will be made too easy for the one
side to win without the use of creative or interesting strategies. We saw in
our sample parliamentary disputation that any answer to a question posed
in debate, other than a direct one, may be called "irrelevant" by an oppo­
nent. However, the problem is to know precisely what is meant here by an
irrelevant answer to a question for in fact, as we saw with the spouse-beat­
ing question and its mates, it may be quite reasonable sometimes to avoid
giving a direct answer to a question. In some instances, to give a direct
answer may be to fall into the questioner's fallacious trap, and in fact many
questions are so aggressively loaded that it would scarcely be fair in logical
games of dialogue to always require a participant to give a direct answer to
any question. On the other hand, giving a totally inappropriate or wildly
irrelevant answer to a reasonable question is something that the rules of
dialectic should discourage and perhaps even in many cases ban. So where
does one draw the line?
Part of the problem here is that it is hard to know how to precisely
define the concept of a presupposition of a question, particularly given that
there are several different types of questions. Belnap (1963) calls a
whether-question a type of question that poses a number of different alter­
natives by means of a disjunction. An example would be 'Is she wearing the
red dress or the green dress?' Belnap then defines the disjunction of these
alternatives as the presupposition of the whether-question. And he defines
a presupposition of the whether-question as any statement implied by any
of the disjuncts. Each of the disjuncts is then called by Belnap a direct
answer. Thus, the statement 'She is wearing the red dress' is a direct answer
to the question, 'Is she wearing the red dress or the green dress?'
According to the approach Hintikka (1976 and 1979), a question is
essentially a request for information. For example, according to Hintikka,
the question, 'What is the colour of Bob's hair?' is equivalent to the
request, 'Bring it about that I know the colour of Bob's hair.' The informa­
tion contained in the question, for example in this case, the proposition, 'I
know the colour of Bob's hair,' is called the desideratum of the question by
Hintikka. Then the presupposition is defined as the existence of the thing
contained within the proposition of the desideratum. In this case, the pre-
 LOGICAL DIALOGUE-GAMES 115

supposition of the question would be There is something such that it is the

colour of Max's hair.' According to Hintikka this is only one perhaps some­
what artificial way of defining the notion of the presupposition of a ques­
tion, but he thinks that it is a good place to start. For Hintikka a direct
answer to a question is one that satisfies the desideratum of the question. In
the present example, a direct answer to the question would be T h e colour
of Bob's hair is red.'
One possible problem with this way of defining 'direct answer' is that
what qualifies as a direct answer appears to be dependent on the state of
knowledge of the questioner. For example, supposing the answerer replies,
"Bob's hair is the same colour as Fred's." Would this constitute a direct
answer? Well, if the questioner happens to know what colour Fred's hair is,
then the desideratum that he know the colour of Max's hair would presum­
ably be satisfied by the answer, and therefore by Hintikka's criterion, the
answer would qualify as direct. However, in other contexts the questioner
might not feel that this answer would fairly constitute a direct answer to the
question. Thus perhaps what counts as a direct answer should be defined
relative specifically to a particular game of dialogue, where relevance is an
issue in that game of dialogue. Managing procedures for irrelevant answers
may become quite difficult to construct however, if we allow other types of
question of a more complex sort, like 'What is the colour of Bob's hair?' or
'Why is chalk white?' Defining the concept of a presupposition, a direct
answer and a relevant answer for these types of questions are substantive
problems of the logic of questions and clearly require much further study.

6. Misconception of Refutation
The term 'fallacy of irrelevance' is applied to many other kinds of fail­
ures of correct argument than lack of subject-matter overlap of component
propositions. Sometimes an argument is said to be irrelevant even if all the
propositions in it are connected by subject-matter relatedness. One such
fallacy occurs where a participant in argument proves some proposition
other than the one he is supposed to prove. True he has proved something
but if he has not proved the proposition that he was supposed to, namely
his thesis, he may have only appeared to succeed in his proof, and may
perhaps have committed a fallacy. The proper name for this type of fallacy
is ignoratio elenchi, as Aristotle called it, or 'ignoring the elenchus.' In Aris­
totle's dialectic the elenchus is the proposition that you are supposed to
prove, so ignoring the elenchus is proving some proposition other than the
116 INFORMAL FALLACIES

conclusion required to be proven. Such a failure of argument may occur

where the proposition you actually prove is similar to, or related to, the
proposition you are supposed to prove, but not identical with it, and far
enough removed that the proof fails.
In ordinary conversation it is not too difficult to appreciate how this
kind of failure of argument functions. For example, suppose that in debate
a participant is supposed to prove the conditional proposition, Tf gun con­
trol were brought in as legislation in the State of New York, then there
would be less crime in the State of New York.' Then suppose that partici­
pant produces a valid argument with the conclusion that gun control legisla­
tion undoubtedly will be brought in for the State of New York in the near
future. And suppose she then appears satisfied, as though the argument is
won. What has gone wrong here, as Alfred Sidgwick (1884) once put it, is
that the journey has been safely performed but the party has gotten on the
wrong train. That is, the argument for the conclusion that gun control will
be brought in, may be quite a strong argument, even deductively valid. But
the arguer has selected the wrong conclusion. She was supposed to prove
the conditional proposition, Tf gun control is brought in to legislation in
New York, then there will be less crime in New York.' So if you look at
what she was supposed to prove in relation to what she did prove, her argu­
ment is deductively invalid. If you look at the argument as a whole, it has
the form A ⊃ (A ⊃ B). Now notice that the problem here is not that of fail­
ure of subject matter overlap. A is indeed related to B, and A is related to
A ⊃ B . That is not the problem. Nevertheless, one might still commonly
say in conversation that the failure of the argument is one of irrelevance.
The conclusion actually established is "irrelevant" to the conclusion sup­
posed to be established, at least in the sense that the former does not estab­
lish the latter by valid argument. Of course, strictly speaking, one might
prefer to say that this is not so much "irrelevance" as just plain invalidity of
argument.
But perhaps we should not quibble over much about precisely what
such a failure should be called. The best term for it is Aristotle's and the
usual translation of this term is the one we have used here, misconception of
refutation. The arguer appears to have refuted her opponent, but has failed,
because the proposition actually proven is really not the thesis to be
proved, nor in a dispute then, would it be the negation of the adversary's
proposition to be proven. Therefore, the move would constitute a failure of
refutation of that adversary's thesis.
 LOGICAL DIALOGUE-GAMES 117

The difficulty with giving a precise analysis of this particular fallacy

comes in when we attempt to define the term 'other than' in the phrase,
'prove something other than the conclusion to be proved'. The problem is
that proof in a dialectical game is not always a one-step affair. In order to
be able to move towards proving her thesis, a participant often needs to
advance more than one single step of argument. She may need to extract
several commitments from her opponent that will eventually enable her to
link up these propositions into a sequence that will move toward her ulti­
mate conclusion. In order to do this she must be given a certain degree of
leeway, since it is not always possible to prove something to the opponent's
satisfaction in one step, or even a very small series of steps. That is indeed
why a game of dialectic allows for several moves, thereby giving each
arguer enough scope to develop a strategy of argumentation. However, at
a global level it is reasonable to require that the sequence of proof should
tend to move towards the ultimate conclusion that the arguer is supposed to
prove. At the local level of one single argument or a few arguments how­
ever, it may be difficult to determine whether a given argument is really
tending towards the ultimate conclusion that should be proven. Therefore
determining misconception of refutation at the local level is often very dif­
ficult for lack of adequate evidence. Once the game is played out to the end
of course, it is much easier to determine whether a given sequence of moves
can fairly be said to constitute an ignoratio elenchi.
The problem remains however that many actual disputes about
ignoratio elenchi in conversation occur at the local level. At this level it may
be hard to say how closely related to the conclusion the given proposition
has to be in order to be said to be relevant.
A well-known example of ignoratio elenchi is given by Copi (1972,
p.86). In this example the prosecution argues at length that murder is a hor­
rible crime instead of attempting to prove that the defendant is guilty of the
crime of murder. As Copi notes, the attorney may even succeed in proving
the conclusion that murder is a horrible crime, but the misconception of
refutation comes in if the attorney expects us to infer from his argument
about the horribleness of murder that the defendant is guilty of that crime.
The conclusion he is supposed to be proving does not follow from the one
he in fact proves. However, in analyzing Copi's example, we have to be
extremely careful. If in fact the attorney has finished his case, and if in fact
his argument does not prove its conclusion or would even be a very weak
argument to establish the guilt of the defendant, then his conclusion, pre-
118 INFORMAL FALLACIES

sumably the one that he did establish — that murder is a horrible crime —
would indeed constitute a failure, and hence possibly a misconception of
refutation.
We would not want to rule out altogether, however, the possibility that
the premiss, 'Murder is a horrible crime' could conceivably have played
some legitimate role in the attorney's strategy to construct proofs for his
conclusion that the defendant is guilty. For example, suppose that part of
the evidence introduced by the attorney was the bizarre behaviour of the
defendant after the time of the alleged crime. The attorney might then
argue that this behaviour was consistent with the commission of a horrible
crime. One could then perhaps see how the premiss 'Murder is a horrible
crime' might legitimately fit into his global argument as part of a strategy
for the conclusion he is supposed to prove, namely that the defendant is
guilty of the crime of murder. In this type of case, the allegation of
ignoratio elenchi might be considerably weaker or even altogether unjus­
tified. So a lot depends on the overall network of the attorney's argumenta­
tion. Once it is finished, we can sit back and more adequately review the
extent to which various premisses in it did or did not play some role in his
argument for the ultimate conclusion that he was supposed to prove.

7. Case Studies of Political Debates

The deployment of the fallacy of many questions is not uncommon in
political debates. It may be a sad comment on the level of parliamentary
debate in many countries of the world to comment on the frequency of this
form of attack, but it can at least be appreciated why resort to this fallacious
strategy of debate is highly tempting to parliamentarians. The basic purpose
of parliamentary debate, within its underlying structure as an adversarial
process, is to attack the position of the opposing party. Indeed, some would
claim that the pursuit of this objective permits the use of any move in argu­
ment, fallacious or not, provided it undermines the opposition's position.
At any rate, some advocates of strong Darwinism of argument in political
debate would allow such moves, provided only they follow the parliamen­
tary rules of order enforced by the speaker in a given assembly.
Whether this strong form of Darwinism is justifiably the right model of
argument for political debate may be questioned. However, it is certainly
true that the main function of political debate is for each party to attempt to
secure the triumph of its own collective position over the opposed positions
of its adversaries. Given this basic objective, every move in argument
 LOGICAL DIALOGUE-GAMES 119

should be directed towards attacking an opponent's position, his set of

accepted premisses or commitments. To this extent at least, the basic objec­
tive of political argument is dialectical and adversarial in nature.
It follows from this objective that basic strategy in political debate
involves the posing of alternatives so that no matter which way the oppo­
nent's position is interpreted or directed, it comes out as a loser in the
debate. The fallacy of many questions is a move of this sort. It poses a ques­
tion that is ostensibly safe. That is, the question has the surface form of yes-
no question — its presupposition is of the form of a tautology, ρ ν lp. How­
ever, beneath the surface the question is force-loaded. That is, no matter
which alternative is chosen by the answerer, he accepts some commitment
that is prejudicial to his own case. So deployment of a many-questions fal­
lacy is precisely the sort of move that seems to be dictated by the principles
of strategy appropriate to the objective of political debate. It forces the
answerer to a losing position in the game of dialogue.
Unfortunately for those who deploy this strategy however, experi­
enced political debaters are rarely taken in by it, and customarily deflect
the question by challenging one or more of its presuppositions, as any
experienced politician knows.
The following illustration of the use of a complex question in political
debate is quite a good one, because the question itself is quite complex. Not
only that, the question is such an aggressive attack on the answerer's posi­
tion that it combines elements of the straw man fallacy and the ad hominem
fallacy within the attack as well. The topic of this particular dialogue, taken
from Hansard {Canada: House of Commons Debates, vol. 127, no.47,
March 12, 1984, p.1994), was government expenditure of funds.7 It arose in
connection with a statement attributed to the minister of justice of the Lib­
eral Party (then in power). The question was put by Mr. Ray Hnatyshyn,
and answered for the Liberal Party by Mr. Marc Lalonde, then minister of
finance.
Hon. Ray Hnatyshyn {Saskatoon West): Mr. Speaker, my supplementary is
directed to the Minister of Finance who, I am sure, had the opportunity of
watching Canada AM when the Minister of Justice was interviewed by
Pamela Wallen this morning. He indicated that he was not in favour of a
policy of the present Government which solves problems by throwing
money at them. He indicated that we cannot continue to go on that way. I
ask the Minister of Finance whether he accepts that point of view and that
criticism of his Government from a Cabinet colleague? Or, if he does not
agree with that statement, does he think it would be more appropriate, if
120 INFORMAL FALLACIES

these Ministers are going to be diverging from Government policy, that

they should do the honourable thing and resign from their portfolios while
they are seeking the leadership of the Liberal Party?
Hon. Marc Lalonde {Minister of Finance): Mr. Speaker, I am very sorry, I
do not watch Canada AM even when I appear on it.
Mr. Epp: I don't blame you.
Some Hon. Members: Oh, oh!
Mr. Lalonde: I have to tell my hon. friend that I fail to understand how he
would be scandalized by such a statement. I made it myself many times in
a lot of areas. The solution to a problem is not throwing money at it.
Mr. Mc Dermid: Why do you keep doing it?
Mr. Lalonde: This is the line we have taken — pursuing a responsible fiscal
and monetary policy, and also, however, showing compassion for people
in need. This has been and will continue to be the policy of this Govern­
ment, contrary to that of the Conservative Party who have singled out the
old, the sick, and the poor, as their targets. And women, as we have dis­
covered recently.
Some Hon. Members: Oh, oh!
Mr. Hnatyshyn reports that he heard the Liberal minister of finance say, on
a radio program, that he was not in favour of the present government policy
of "throwing money at problems." Already in advancing this accusation,
Mr. Hnatyshyn could be questioned concerning whether the allegation is a
form of straw man criticism. For it is dubious whether his report of the
alleged statement by this minister fairly represents the Liberal Party's posi­
tion or the government's policy. Expressed in such prejudicial terms, it is
highly unlikely that such a description fairly represents the Liberal position,
and Mr. Lalonde's reply confirms these doubts. Even so, now presupposing
this statement that the government has a position of "throwing money at
problems," Mr. Hnatyshyn's yes-no question first poses the alternative of
whether Mr. Lalonde agrees with the position. Then he offers the remain­
ing alternative: if Mr. Lalonde does not agree with the position, when is he
going to resign?
The yes-no question offers two alternatives. First, if the answerer
agrees to the "government position" of "throwing money at problems" then
of course he is guilty of irresponsible financial management and of support­
ing an indefensible position. But on the other hand, if he disagrees with his
(alleged) own party position, then he is inconsistent on the grounds of
 LOGICAL DIALOGUE-GAMES 121

divergency from his own government policy. This second prong of the yes-
no question is a form of circumstantial ad hominem attack. The allegation
is that the answerer's position is inconsistent.
Mr. Lalonde's reply is, of course, to rebut the presumption of the ques­
tion by saying that he has claimed many times that the solution to a prob­
lem is not throwing money at it. He disclaims knowledge of the statement
on the program by the minister of justice. Finally, he rebuts the prejudicial
presumption of the question even more positively by asserting that the pol­
icy of the Liberal Party is described as "responsible fiscal policy" and
"showing compassion for people in need." This description is, of course, an
oppositely prejudicial account of the Liberal position to the negative
description presumed by Mr. Hnatyshyn's question.
The dialogue has achieved nothing by way of clarifying the position of
either party on the issue under dispute. So Mr. Hnatyshyn's attack, based
on a report of someone's statement poorly documented in specifics of
phrasing, comes to nought.
Mr. Lalonde failed to give a direct answer to Mr. Hnatyshyn's ques­
tion. However, that is no fault of his reply in this instance because the ques­
tion itself was fallacious. Mr. Lalonde's rebuttal of the presumption of the
question appears to be a reasonable reply. In fact, we might say that, in the
circumstances, it is a proper sort of reply to the question.
Another very interesting example of misconception of refutation
occurs in the following case study of parliamentary debate from the House
of Commons Debates of Canada, Volume 124, no.270 (First Session, 32nd
Parliament, December 9, 1981, page 13906).
In this debate Mr. Friesen, the first speaker, raises a point of order
concerning a rule in Beauchesne (the rule book for the House of Commons
Debates in Canada). The rule stated in Beauchesne is that a member of Par­
liament must address the House orally and not read from a written, previ­
ously prepared speech. This does not mean that the member must not use
notes, nor does it mean that he must give a perfectly original speech. He is
quite free to memorize and then present orally any material that he wishes.
The rule simply requests members not to read off prepared notes from a
written page, in a monotonous lecture. Rather, the speaker is supposed to
keep his eyes on the persons he is addressing and at least give the appear­
ance of speaking rather than simply reading from notes.
Mr. Blaker, the Acting Speaker, then wonders whether all the mem­
bers would really like this rule to be applied strictly to everyone. No doubt
122 INFORMAL FALLACIES

many of the members do tend to take a good deal of their speaking material
from prepared notes.
Then at that point in the debate, another member, Mr. Taylor, rises on
a point of order and accuses Mr. Blaker of reading what somebody else
wrote! That is, Mr. Taylor is accusing Mr. Blaker of reading somebody
else's speech, in effect, accusing him of plagiarism. He then adds that this is
contrary to the rule and suggests that he close his notes and then try to give
the speech.
In responding, Mr. Blaker points out quite correctly, that Mr. Taylor's
point is, as he puts it, entirely irrelevant, and that nothing the rule says is
contrary to a member reading the collective works of Shakespeare word for
word. The issue is not the identity of the author that the member is refer­
ring to, but whether he is reading a speech, not whether that speech has
been written by himself or someone else. Here then is the full text of the
debate.
Mr. Friesen: On a point of order, Mr. Speaker, Beauchesne at page 101,
rule 309, reads as follows:
It is a rule in both Houses of Parliament that a Member must address the
House orally, and not read from a written, previously prepared speech.
I do not mind if the hon. member has copious notes. We know he can
read. We would like to hear an extemporaneous speech from him rather
than a parliamentary secretary's or bureaucrat's speech.
The Acting Speaker {Mr. Blaker): This particular occupant of the Chair has
had occasion to deal with this point of order before. The hon. member for
Surrey, White Rock, North Delta (Mr. Friesen) is perfectly correct, mem­
bers are encouraged to avoid reading speeches verbatim. The difficulty
that the Chair faces whether it be myself or some other occupant, is that if
the House wishes to instruct the Chair that this rule is to be followed, the
Chair will instantly and with great pleasure do so, but it will apply equally
to all members of the House at all times.
I would recommend that the hon. member give some thought to whether
indeed that is the rule he would like applied to members on all sides. If
that is the case, I am sure the Chair will be notified of that. Aside from
being astounded and delighted, the Chair would be pleased to put that into
effect immediately. I suggest that some consideration be given in caucus to
that matter before the hon. member asks me to make a decision. If the
hon. member wants the rules applied, I will apply them.
Mr. Taylor: I rise on the point of order, Mr. Speaker. The hon. member is
simply reading what somebody else wrote. That is completely contrary to
 LOGICAL DIALOGUE-GAMES 123

this rule. I ask him to close his notes and give his speech. He cannot do it.
He is reading word for word. Somebody else wrote that speech and he
knows it. Also, they vote for closure and then try to waste our time. Why
did you vote for closure you bunch of hypocrites?
Some hon. Members: Oh, oh!
The Acting Speaker (Mr. Blaker): Order, please. The point is entirely
irrelevant. If the hon. member wants to read the collected works of
Shakespeare word for word ....
Mr. Taylor: Shakespeare didn't write that trash!
Some hon. Members: Hear, hear!
The Acting Speaker (Mr. Blaker): The point would still remain that the
issue is not the author of what an hon. member may refer to but rather
whether or not he is reading a speech. I can add no more to what I have
already said to the earlier members who rose on the same point of order.
Mr. Taylor's objection here is a misconception of refutation. What he
should have objected to was the member's reading the speech instead of
addressing the House orally. But instead he objected on the issue that the
member who read the speech was not the author of the speech. This objec­
tion, although it may be subject-matter related to what the rule forbids, is
nevertheless quite a distinct proposition. What Mr. Taylor should have
objected to was that the opposition member was not addressing the House
orally and not raised the issue of whether the member had in fact himself,
written this previously prepared speech or not. Thus, Mr. Blaker, in his
reply to Mr. Taylor's objection, quite correctly uses the word "irrelevant"
in his rebuttal that Mr. Taylor's point is "entirely irrelevant." But evidently
what is meant here by 'irrelevant' is that a misconception of refutation in
Aristotle's sense has taken place. Mr. Taylor's objection is not entirely
irrelevant to the issue that he should be addressing himself to in the sense
of being entirely subject-matter disjoint. His objection does indeed share
some subject-matters with the statement he is supposed to be arguing
about. His attempted refutation is relevant (related to the issue). But it is
still a misconceived refutation and therefore "irrelevant" in the Aristotelian
sense.
Part of the problem of this particular debate appears to be related to
the fallacy of equivocation. For there appears to be a certain ambiguity in
how the phrase "reading a prepared speech" may be taken. The other prob­
lem is that of defining "relevance" in terms appropriate to the particular
124 INFORMAL FALLACIES

context of argument. We have already seen that this problem is pervasive in

political debates because the precise objectives of such free-for-all adversa­
rial contests are very hard to pin down. No doubt the participants them­
selves could get into a lot of trouble if they tried to decide what should con­
stitute "relevance" as they see it.
Let us conclude by summing up the main problems we have found in
attempting to apply any model of reasonable dialogue to realistic debates
and fallacies.
Even with simple yes-no questions only allowed in dialogue, we can get
into a lot of trouble. There remains the problem of determining when a 'No
commitment' answer is reasonable or permissible. Moreover, even the
notorious spouse-beating question has the form of a yes-no question! Yet if
one directly answers this question, one is undone. In this case, the question
really is fallacious. So we have already got quite difficult problems to solve
in connection with the yes-no and why-questions of games of dialogue.
Defining 'direct answer' for other more complicated or richer types of ques­
tions is a more advanced problem.
What this means for the evaluator of dialogue is that one should be
careful to realize that ruling on what should constitute a "direct answer" to
a real question may require a good deal of judgement of the particular cir­
cumstances of the dialogue, the questioner, and the answerer. To orient
this type of inquiry however, it is always good to begin with a consideration
of the question-answer sequence as a regulated game of dialogue. We need
to ask: what is the purpose or objective of this particular game of dialogue?
We need to at least try to formulate what could be analogous to the win-loss
rules of the game. In other words, it is a good idea to think of the dialogue
being analyzed as an instance of one of the models of games of dialogues we
have studied. In many cases, this task is not easy, and requires a good deal
of judgement and empirical information. For example, suppose one is
attempting to analyze a parliamentary debate, to see if an answer to a
member's question should be reasonably judged as relevant. One needs to
ask: what is the objective of parliamentary debate, or at least, of this par­
ticular debate? This question may be somewhat like an essay topic in polit­
ical science! But it is, at any rate, worth formulating some general state­
ment of the objectives of the dialogue in order to frame standards for the
reasonableness of what should count as a direct answer to a question in the
course of the dialogue.
Many of the same kinds of problems are implicit in defining the con-
 LOGICAL DIALOGUE-GAMES 125

cept of presupposition of a question. With yes-no questions and whether-

questions, the concept of presupposition can be clearly defined in a rela­
tively noncontroversial way. However, with questions not of these forms,
judgement may be required to reach agreement on what should fairly count
as the presuppositions of a particular question.
Finally, any theory of dialogue that hopes to be applicable to the fal­
lacies should be able to contend with the ad ignorantiam fallacy in the man­
agement of question-asking. To contend with all these various problems of
dialogue, we turn to a new kind of game of dialogue, different from any of
the theories so far advanced.

8. A Game with Dark-Side Commitments

If dialectic is to be made a model of argument best suited to its prag­
matic role as a method of argument and fallacy analysis, some improve­
ments could be suggested. According to Hamblin, games of dialogue are
"information-oriented," and the purpose of the game is for the participants
to exchange information. Hamblin does not give specific rules for telling us
when this objective has been achieved. What seems lacking is a precise for­
mulation of what constitutes "win" or "loss" of the game.
Hintikka's conception of dialectic is quite precise in this regard. The
player wins who first accomplishes his objective of deducing his thesis, by
the inference-rules of the game, from his opponent's concessions. But one
point where Hintikka dialectic could be queried as a model for the fallacies
is the rule for answering questions. According to Hintikka (1979) a player
may take the option of not committing himself one way or the other in
answer to a "yes-no" question. But if he does so, the negation of the pre­
supposition of the question is added to his list of commitments. This would
mean, for example, that failure to answer the spouse-beating question
results in automatic commitment to the proposition, There is no time at
which I have stopped beating my spouse.' Is this a best solution or not? It
depends on whether this reply exonerates the questioner or commits
him(her) to still continuing with the practice of spouse-beating.
Because of the need to study different varieties of dialectical struc­
tures, I have proposed two types of innovations in Walton (1983). First, we
may distinguish between two compartments of a player's commitment-store
— his "light-side" commitments, known to him and the other player, and
his "dark-side" commitments, not known to any of the players. However
dark-side commitments can be guessed or inferred with plausibility, and
126 INFORMAL FALLACIES

therefore may play some part in a player's strategy. The dark-side position
of a player is a definite set of propositions, written on a slate or piece of
paper, but the players do not know themselves what each player's dark-side
set is. As the game proceeds, however, the dark-side propositions begin to
come over to the light side. How does this take place?
If a questioner replies 'No commitment p' to a 'yes-no' question, but in
fact ρ is in his dark-side commitment-store, then ρ is transferred over to the
light side. In an instance like this, where the player is committed to p, but
at the same time has replied 'No commitment p' we say that his position is
ambivalent. Any ambivalent position must be challenged by the other
players, and the ambivalence must be resolved or the ambivalent player
loses the game. Hence in these new games of dialectic, a player is "gently
forced" to make commitments in answer to questions. But he is not so
strongly forced that he must accept some proposition that isn't really part of
his position or consistent with his position as far as he knows.
This innovation is an exciting one, for as we have seen with many of
the fallacies, an arguer's position is most often dimly known by himself or
his critics, and only begins to become articulated through the course of the
argument itself. By this structure, dialectic can be a dynamic and informa­
tion-oriented process of argument that can handle realistic situations of dis­
putation and criticism where "missing premisses" need to be filled in by fair
criticism.
Hence this new model of dialogue is especially applicable to the prob­
lem of enthymemes and to the problem of fairly adjudicating ad hominem
disputes. However, we reserve the study of these problems for later chap­
ters. Our main concern of the moment is the management of reasonable
procedures for question-asking in dialogue. This new game of dialogue
starts at a simple level of discussion by only allowing yes-no questions.
However, in later chapters we will add more complex versions that also
include why-questions.
The following game of dialogue has the special feature of having the
commitment-sets of each player divided into two partitions. One side of the
commitment-set is on full view to both players at all times (the light side).
The other side, called the dark side, is a set of propositions contained in the
player's slate of commitments, but not in view of either player. 8 The dark
side set of propositions is a set that definitely exists, but neither player can
see it.
Another distinctive feature of this particular game, which we will call
 LOGICAL DIALOGUE-GAMES 127

the game ABV, is that only one kind of question-move is permitted. The
players may ask only yes-no questions. They may not ask other kinds of
questions. In the B-series of games, the players can ask only why-questions.
In the C-series of games, the players can ask both kinds of questions.
The purpose of each player in an ABV game is to prove his own thesis
as the conclusion of a valid argument from premisses that are commitments
of the other player. Each player has a proposition designated as his thesis
(or conclusion) established at the outset of play. To win, each player must
prove his thesis from the other's commitments. The commitments represent
the propositions that a player accepts, either before the game begins or dur­
ing the course of play. A main function of the rules is to define how such
commitments are incurred at each move.
There are two players, called White and Black. The game may be
defined as a set of moves. A move is defined as an ordered pair composed
of a locution and a player. Where L is the set of locutions and Ρ the set of
players, the game of dialogue may be defined as (L X P) n , the sequence of
η locution acts (locution events). Where m0 is the first move — by conven­
tion, White always makes the first move — the set of moves, m0, m1' ...,
m , terminates in the last move mn . At any move mk, (mn >mk ≥ m0) there
will be a next move in the game.
A game of dialogue is defined as a triple < L , P, K > where  is a set
of rules that defines the legal dialogues (set of permissible moves) for that
game. For each particular game, there must be a set of locution rules, a set
of dialogue rules, a set of commitment rules, and a set of win-loss rules
(strategic rules). The locution rules define the locutions permitted as admis­
sible moves. The dialogue rules define the permissible response (next
move) for any move. The commitment rules define the alteration of com­
mitments consequent upon each move. Each player has a commitment-set,
a set of propositions that he agrees to accept at the outset of the game. It
may be an empty set. The win-loss rules define which sequences of moves
count as win, loss, or draw for each player. In a zero-sum game, one player
wins if and only if the other loses. There will also be a set of logical rules
defining what counts as a logical consequence of any set of propositions.
128 INFORMAL FALLACIES

The Game ABV

Locution Rules
(LI) Statements: Propositional variables, A, B, C, ..., are permissible
locutions, and truth-functional compounds of propositional vari­
ables.
(L2) Withdrawals'. 'No commitment A' is the locution for withdrawal.
(L3) Questions: The question 'A?' asks, 'Is it the case that A is true?'
(L4) Argument Queries: The query 'A1 ,A2 , ..., An , therefore B?' asks
'Do you accept  on the premisses A1 ,A2 , ..., A n ?'

Dialogue Rules
(Dl) White moves first (at m0) and asks a question.
(D2) A question 'A?' asked by one player at any move mk must be fol­
lowed by the other player at the next move mk+1 making exactly
one of the following three types of moves: (i) 'Statement A', (ii)
'Statement IA', (iii) 'No commitment A'.
(D3) When one player (the responder) at m k + 1 responds via (D2) to a
question asked by the other player at m k , then at m k + 2 the
responder must ask a question.

Commitment Rules
(CI) After a player makes a statement A, it is included in his commit­
ment-store.
(C2) After a player withdraws a statement A, it is deleted from his
commitment-store.
(C3) If a player makes an argument query 'A1 , A2 , ..., A n therefore
B?' and all the statements A1 , A2 , ... A n (for finite n) are in the
commitment-set of the player queried, and the argument queried
is a substitution instance of a logical rule of the game, then the
player queried must reply 'Statement B' in answer to the query.
(C4) When a player answers 'Statement B' in accord with the rule
(C3) above,  cannot henceforth be deleted from his commit­
ment-set.
(C5) If a player states 'No commitment A' and A is on the dark side of
his commitment-set, then A is immediately transferred to the
light side of his commitment-set.
 LOGICAL DIALOGUE-GAMES 129

Win-Loss Rules
(Wl) Both players agree in advance that the game terminates at m for
some fixed, finite n.
(W2) A player wins at m k when the other player makes a statement A
at m k , where A is the first player's thesis.
(W3) If neither player wins by the last move m , the game is a draw.

Logical Rules
The set of logical rules can be any set of valid rules for classical propos-
itional calculus, complete or incomplete. However, both players must
specify such a set, and record their agreement in advance of play.

This game is similar to the Hamblin game (H) in having Locution

Rules, Dialogue Rules and Commitment Rules. It is, however, a special
case of a Hintikka game because it has Hintikka-style precisely formulated
win-loss rules. It is also dissimilar to (H) in that it lacks why-questions.
ABV is non-cumulative, because a player can, at appropriate
junctures, delete commitments by (L2). However, rule (C4) does mean that
there is one kind of restriction on retractions. If one player proves by an
argument query type of move that the other player must be "logically" com­
mitted to some proposition  because it follows from his commitments
directly by a logical rule, then that second player cannot subsequently
retract his commitment to B. In this way, there is a sort of "partial cumula-
tiveness" inherent in the game ABV. Players clearly cannot retract all their
commitments. But except for this one type of instance, they are otherwise
free to retract commitments by means of (L2).
Readers of Walton (1984) will observe that ABV is like the game CBV
in general outline, except that CBV allows why-questions as well as yes-no
questions. Moreover, there are some smaller differences of formulation and
detail. For example, the mechanism for argument queries is organized a lit­
tle differently in CBV. These smaller differences are not meant to be too
significant however, and are more meant to be improvements of formula­
tion. In general then, ABV is meant to be a special case of CBV, a simpler
game that lacks the why-question feature of CBV. The rules for CBV will
be given in the next chapter.
In ABV, the dialogue rules are meant to regulate the pattern of moves
made by the players. Each player must take turns asking and answering
130 INFORMAL FALLACIES

questions. First one asks and the other answers. Then at the next move, the
former answerer becomes the questioner. In this regard, we could say that
the dialogue represents a symmetrical game, where the players "exchange
roles" at each pair of moves.
In a non-symmetrical game representing a fragment-game of ABV,
there could be just one questioner and the other player always has the role
of answerer. In this non-symmetrical type of game, the questioner wins if
he proves his thesis on the basis of the answerer's commitments in an
agreed-upon finite number of moves. Otherwise the answerer wins.
One of the main stumbling blocks to studying the fallacy of many ques­
tions and related problems of question-asking was that the purely syntactic
or semantic definitions of 'presupposition' did not do justice to the reality
that a loaded or prejudicial question in one context might be a reasonable
question in another context. For example, if addressed to an acknowledged
spouse-beater, the spouse-beating question could be reasonable. In another
context, it would be fallacious. However, in ABV we can define a pragma­
tic notion of 'reasonable presupposition' of a question in the context of a
game of dialogue. A presupposition of a question is reasonable at some
point in a dialogue if at that point the proposition expressed as the presup­
position is contained in the light-side or dark-side commitment sets of the
player to whom the question is asked.
This definition of reasonable presupposition is, however, a narrow
one. A wider definition should include not only propositions a player does
accept, but also propositions he can accept without weakening his own posi­
tion in the argument. This wider definition involves notions of strategy in
argument as defined in Walton (1984). We will develop these concepts of
strategy in the game CBV in the next chapter. At any rate, for the present,
we have developed a new basis for further studies of the practical problems
of the management of questions posed in this chapter.

NOTES

1. The word 'dialectic' stems from the ancient Greek term for a discussion by means of ques­
tion and answer. We use the terms 'game of dialogue' (Hintikka) and 'game of dialectic'
(Hamblin) interchangeably. As we use the term 'dialectic' it is not meant to have any connection
with its use by Hegel or Marx as a term for a process of historical development.
2. Variants on Hamblin games can be found in Mackenzie (1979) and (1981). A quite differ­
ent sort of framework for dialectical games is to be found in Rescher (1977).
 LOGICAL DIALOGUE-GAMES 131

3. The theoretical structures needed to construct dialectical games are set out by C.L.
Hamblin, 'Mathematical Models of Dialogue', Theoria, 37, 1971, 130-155.
4. A fuller analysis of the ad ignorantiam argument is given in John Woods and Douglas Wal­
ton, 'The Fallacy of Ad Ignorantiam', Dialectica, 32, 1978, 87-99.
5. As Peter Geach pointed out to me however, we should be clear that there need be nothing
wrong per se with asking a conjunctive question like, "Did you buy the eggs and clean the rug?"
It is only in certain dialectical circumstances that multiple questions can go wrong.
6. A fuller analysis of the fallacy of many questions is given in Douglas N. Walton, 'The Fal­
lacy of Many Questions', Logique et Analyse, 95-96, 1981, 291-313.
7. I would like to thank Fong Kim Ng for drawing this debate to my attention.
8. This innovation was first suggested in a meeting of the Logic Seminar of Victoria Univer­
sity of Wellington (New Zealand) by Max Cresswell in March, 1983.
 CHAPTER 5: ENTHYMEMES

One of the most immediate and serious problems of any attempt to

apply logic to real argumentation is that in arguments as they are really
stated, so much is left out. If you try to give an analysis of an extended dis­
course that expresses an argument, the initial problem is that there will be
many gaps in the chain of argument, constituted by missing propositions
that are plausibly meant to be premisses by the arguer, but that have not
been explicitly stated in so many words. The first job of reconstructing any
argument is to evaluate the place of missing premisses.
As we saw in chapter 1, an argument that has a missing premiss is,
according to tradition, called an enthymeme, or enthymematic argument.

1. The Tradition of Enthymemes

According to traditional logic texts and manuals, and enthymeme is an
argument with one or more enthymematic premisses. An enthymematic
premiss is a premiss not explicitly stated, but tacitly presumed in the argu­
ment. The argument, 'All men are mortal, therefore Socrates is mortal,'
according to the traditional doctrine of enthymemes, tacitly assumes the
additional premiss, 'Socrates is a man'. In this case, when you add the
missing premiss, the argument comes out deductively valid in classical first-
order logic.
The tradition stems from Aristotle. In Analytica Priora, Book II (70a
10), Aristotle defines an enthymeme as a syllogism that starts from a gener­
ally approved proposition, e.g. 'The beloved show affection.' In Rhetorica
(1357a 18), Aristotle writes that an enthymeme is a shortened syllogism —
if one of the premisses is a familiar fact, there is no need (for purposes of
persuasion) to mention it. The rhetorical persuader, we are told, should
avoid reasoning that is too hard to follow because of its length.
The idea seems to be that if an argument is deductively invalid as it
stands, but is "fairly close" to a deductively valid argument, then you can
"plug the loop-hole" and make it into a valid argument. One problem with
this, however, is that there may be different ways to plug the loop-hole. We
134 INFORMAL FALLACIES

could have put in 'Socrates is a man and Plato is a man' and the argument
would also be rendered valid by that addition.
The doctrine of enthymemes would seem to suggest the rule: always
add the weakest premiss needed to make the argument valid. This will not
do, however. The proposition 'Socrates is mortal' is weaker than the propo­
sition 'All men are mortal,' but if the argument given were 'Socrates is a
man, therefore Socrates is mortal,' the correct enthymeme would presuma­
bly be the latter rather than the former proposition.
It is not easy to say what could be meant by 'weaker than' as a relation
here. In classical logic 'Socrates is mortal' implies 'Socrates is mortal or all
men are mortal'. But is the second proposition "weaker than" the first?
Possibly something like Parry's (1933) notion of analytic implication could
be useful.
Even if we would bring to bear a satisfactory account of the 'weaker
than' relation, two problems remain. First, the missing premiss wanted may
not be the weakest proposition, but rather the "most plausible" one of the
various ones that would be sufficient to make the argument valid. For
example, the well-known principle of charity recommends adding the
missing premisses that the arguer most plausibly had in mind from what we
know of his position and the context of the argument. But how to select the
"most plausible" proposition from the multitude of sufficient candidates
available? I don't see any obviously correct general logical procedure for
carrying out such a selection.
The standard doctrine of enthymemes, as it is to be found in current
logic texts, is reviewed by van Eemeren and Grootendorst (1983, chapter
6). They quote one leading text as stating that the principle of charity gov­
erning enthymemes is that one should try to make the argument valid and
its premisses true — insofar as this is possible. However, van Eemeren and
Grootendorst (p.125) point out that the qualification "insofar as this is pos­
sible" is crucial. For sometimes making the argument valid can only be
done at the cost of using a premiss that is patently untrue. How one is to
apply the principle of charity, therefore, is a matter that is problematic and
unclear. In this light, the question of enthymemes should be reconsidered.
Charities have sometimes been criticized as a paternalistic second-best
type of solution when one should really help the needy to help themselves.
In this light, should we really be asking the enthymematic arguer what he
means to say, or co-operatively him to "say it better" rather than just plug­
ging in what we — the critics — think is the most plausible proposition?
 ENTHYMEMES 135

Some would call the two approaches of charity and the weakest loop-hole
very dangerous from a strictly logical point of view. Strictly speaking, you
should never assume that your opponent in argument, or one whose argu­
ment you are prepared to criticize, has made assumptions that she has not
clearly stated.
Philosophers have sometimes warned us of the enthymematic p l o y : you
can always make a good argument from a bad one by filling in some missing
premisses. But once you start fiddling with your argument, it is — strictly
speaking — a different argument. You can try to defend your argument by
replying to the discovery of a loop-hole: ' O h , well of course I meant to say
that as well." But did you? Simply calling the missing bit an "enthymeme"
is too easy a way out to always allow. There is danger of logic becoming
unstuck here, in a sea of fluctuating premisses.
There is also the danger of making every argument into a valid argu­
ment. For example, you can make inductive arguments into deductively
valid arguments by adding ceteris paribus clauses or closure conditions to
the premisses. But this strategy has often seemed specious. You may be
really only adding a premiss that can only be applied by using or presuppos­
ing some inductive technique. The suspicion is that by rendering the argu­
ment deductively valid you have begged and obscured the questions of
whether and why it is any good. Could it be that the whole doctrine of
enthymemes is pernicious as a part of logic? It seems to make a certain
amount of sense to just ask the arguer: "Is this missing premiss what you
want to say?" If so, determining enthymemes is a matter of the psychology
of belief, or a question of asking for additional information, not a matter of
logic. It may be that Aristotle had something like an empirical approach in
mind in Rhetorica (1357a 20) where he gives the following example of an
enthymeme: "... to show that Dorieus has been victor in a contest for which
the prize is a crown, it is enough to say 'For he has been victor in the Olym­
pic games,' without adding 'And in the Olympic games the prize is a crown,
a fact which everybody knows." Here "what everybody knows" could be
cashed out as some empirical datum, perhaps. But such an outright empiri­
cal approach is not quite fair to the traditional doctrine of enthymemes. For
the idea behind the doctrine seemed to be that somebody's argument might
be committed to some unstated, but clearly necessary and relevant assump­
tion, which should therefore be counted in to the argument, even if disav­
owed, perhaps when later recognized as open to criticism.
Van Eemeren and Grootendorst (1983, p.141) seem to agree on this
136 INFORMAL FALLACIES

point. They reject the idea that the critic of an argument, when explicitizing
unexpressed premisses, must try to select propositions that the speaker
actually believes. Apart from the practical difficulties, e.g. in written texts,
they believe this approach to be fundamentally mistaken. It is their view
that the critic of an argument must try to establish what propositions the
arguer is committed to, with the assistance of valid argument forms and the
context of argumentation.
The speaker can also be held to statements to which he has committed
himself implicitly, so that in principle he is also obliged to defend such
statements. Whether these statements coincide with what the speaker 'ac­
tually' thought or subsequently comes to believe is irrelevant (p. 141).
The problem then is to get the requisite notion of commitment
required to make some sense of this doctrine without (a) giving the defen­
der of the argument complete autonomy to change his argument by filling
in the loopholes any way he decides to, during the course of the argument,
or (b) giving the critic the paternalistic power to fill them in whenever he
wants and however suits his needs for criticism. Possibly (b) is a worse
danger than (a) in many cases. So a sensible and useful theory of
enthymemes should avoid acquiescing in (b) too heavily, yet without com­
pletely giving in to (a). It is a question of justice in what you can fairly or
reasonably assume in an argument. This being the case, the framework of
logical dialogue-games could be the best place to turn.

2. The Objectives of Dialogue

According to the account of enthymematic arguments given by van
Eemeren and Grootendorst (1983, p. 141), there are three conditions that
must be met by an explicitized, unexpressed premiss: it must be (1) an
informative statement that is (2) a commitment of the speaker, and (3)
makes the argument valid when added as a premiss. These three conditions
are filled in by the listener, according to van Eemeren and Grootendorst,
by the principles of Gricean conversation theory. The preparedness of the
listener to fill in a missing premiss is an instance of the Gricean principle of
co-operation — by contributing to the resolution of the dispute, the listener
is acting in a co-operative manner.
However, from the normative point of view of logical dialogue-games
outlined in the previous chapter, a somewhat different approach to supple­
ment the perspective of conversation theory could be suggested. We recall
 ENTHYMEMES 137

that according to Hamblin, dialogue is information-oriented in the sense

that the purpose of the participants should be to exchange information.
Hence both the ideas of 'informative statement' and 'commitment of an
arguer' are familiar in Hamblin dialogues.
Hamblin (1970) argued that the best way to study fair and unfair moves
of argument is to set up dialectical games (systems) that model discussions
or dialogues, the natural environment and context of criticisms and fallacies
as they have been traditionally conceived. As he sees it, dialectical systems
can be pursued descriptively or formally. The descriptive study looks at
rules and conventions of real discussions like parliamentary debates or legal
cross-examinations. The formal approach involves the construction of sim­
ple but precise systems where moves are regulated by rules that can be
clearly stated even if they may not necessarily be realistic. These formal sys­
tems will then have formal properties that can presumably be compared to
interesting sequences of realistic discussions and thereby throw some light
on the latter by modelling them.
Hamblin (1971) defines a dialogue as a set of locutions, L, and partici­
pants, P. By a dialogue of length n, he means a member of the set (Ρ Χ L) n
of sequences of η locution-acts. A locution-act is a member of the set Ρ X L
of participant-locution pairs. Next, a set of rules is added which defines
within a dialogue D a set of legal dialogues K. A system is a triple (P, L, K).
Hamblin's formal constructions are concerned with possible definitions and
properties of K.
A Hamblin game of formal dialectic then must involve a set of
"players" and "moves" made by these players. A third key ingredient is the
commitment-store of each player. Commitments are not beliefs of the
players, but operate approximately like the real beliefs of an arguer. How­
ever, psychology is not the purpose of constructing Hamblin games, and we
are advised to think of a commitment-store, strictly speaking, more along
the lines of a set of statements written down by each player on a slate that
he possesses. As we have seen, the rules of a Hamblin game add to or sub­
tract from the commitment-stores of the players, and how this modification
of the stores takes place is the key to modelling the fallacies.
Hamblin considers the requirement that commitment-stores should
always be internally consistent (p.257) but rejects it, at least as a universal
requirement on dialectical systems because it is an ideal of 'rational man'
not always met with (p.263). He is also inclined to reject deductive closure
of commitment-stores as a universal requirement, but (p.264) feels that
138 INFORMAL FALLACIES

"certain very immediate consequences" of a commitment may also be com­

mitments. Both requirements are matters of "regulation in a given system"
(p.264).
Hamblin (1970, p.265-8) has designed one particularly basic game we
may call (H), with the purposes of realizing a concept of argument and
modelling some of the traditional fallacies. There are two participants,
White and Black, who take turns making moves. The types of moves
allowed involve the asking and answering of questions. Hamblin (p.265)
formulates five rules that demarcate permissible locutions. Capital letters S,
T, U, ... are variables for statements.
(i) 'Statements S' or, in certain special cases, 'Statements S, T'.
(ii) 'No commitment S, T, ... X', for any number of statements S, T,
... X (one or more),
(iii) 'Question S, T, ... X?', for any number of statements (one or
more),
(iv) 'Why S?', for any statement S other than a substitution-instance
of an axiom.
(v) 'Resolve S'.
The language of (H) is propositional calculus, or any other "suitable" sys­
tem with a finite set of atomic statements. Each participant has a commit­
ment-store, a set of commitments that contains the axioms for the lan­
guage. There are two types of questions that a player can ask, (iii) or (iv).
However Hamblin notes that two simpler games could be built by deleting
one or the other of these rules and keeping the remaining four.
Following van Eemeren and Grootendorst's suggestion, we could use
Hamblin's framework to work towards an account of enthymemes. In (H)
you could rule that a proposition is a fair assumption to make as a missing
premiss in a player's argument if that proposition is in that player's commit­
ment-set. This does not uniquely define an enthymeme for a given argu­
ment, but it seems to narrow them down in the right way. At least it
excludes the statements that an arguer has not accepted or is not committed
to.
There does remain a problem, however. If I am constructing an argu­
ment against your argument, what am I allowed to assume as my "en­
thymemes," your commitments or mine? Or to be a proper enthymeme,
must a proposition be in the intersection of your and my commitment-set?
It seems hard to definitively rule on this question because it is not precisely
 ENTHYMEMES 139

formulated what counts as a win or loss of a Hamblin game if it comes down

to a contestive dispute.
Hamblin writes (1971, p.137) "that the purpose of the dialogue is the
exchange of information among participants." What precisely counts as
"exchange of information" is not defined, but Hamblin's general presump­
tion that games of dialogue should be "information-oriented" (his term,
p. 137), does affect how he designs (H), and that affects how arguments are
analyzed in (H). For example, Hamblin suggests that there is no point in
asking a question if one is already committed to one of the answers (p. 137),
and the rules of Hamblin games tend to reflect this information-oriented
design of rules for questioning.
The problem here is that a game of dialogue is partly co-operative and
partly contestive. In the context of fallacies and criticisms, the objective of
dialogue should, to some extent, be treated as adversarial. The objective of
each participant is to prove something to the other.
If the structure of formal dialogues are to reflect the practices of realis­
tic dialogue-interchanges of proving and refuting arguments, some notion
must be brought in of a participant adopting a strategy — a hypothetical
sequences of moves — in order to fulfill his objective in the disputation.
The answerer's objective, let us say, is to prove his thesis A to the ques­
tioner. In a dispute, the questioner's objective is to prove the opposite of
Τ . Hence the answerer knows that the questioner is strongly committed to
resist commitment to Τ . If the answerer tries to "prove" in one step, by
taking a commitment of the questioner as premiss, then one of two things
will happen. If there is in fact such an S that is a commitment of the ques­
tioner and S implies Τ , then the answerer wins the game if the questioner
cannot retract any of his commitments. If the game allows retractions, the
questioner is most likely to simply retract his commitment to S, providing
he sees that A, the thesis of his opponent, is a direct consequence. Of
course there may be no such S available in any event. Generally, if the par­
ticular game in question is to be of any practical interest, there will be no
such S directly available to the answerer. What then is the answerer to do?
The answer is that he must adopt some sort of strategy. Typically in
practice, the answerer will not know how strongly his opponent is commit­
ted to some of the statements in his commitment-store as opposed to
others. But in order to adopt a working strategy to fulfill his objective, it
would be useful if he could roughly order the statements he proposes to use
as premisses according to how likely he thinks it to be that his opponent will
140 INFORMAL FALLACIES

accept them. He must ask himself "Which one of the two propositions is my
opponent more likely to think plausible or at least congenial to his own
position?" By asking himself a series of such questions, he may be able to
organize all the statements he might eventually find useful as premisses into
different levels of acceptability. Putting his proposition to be proved, say
Τ , at the lower bound of the order, he should then proceed to construct a
line of proof that starts as close to the upper bound of the order as possible
and proceeds deductively towards the lower bound. That procedure is the
general form of a best strategy for the answerer.
But the problem is that these linkages are loose, and one needs to
know how the purpose of a game, its information-orientation, specifically
affects the strategy of the players.
The Hintikka games of dialogue do not share this open-ended quality
of win-loss determination. Quite to the contrary, the win-loss rule for a
Hintikka game of logical dialogue is precisely defined. A player wins if, and
only if, he deduces his own thesis by the rules of the game from his oppo­
nent's commitments. In this regard, a Hintikka game is precisely regulated.
It is quite clear how the objective of each player is set. And therefore, in
general outline it is possible for each player to plan a strategy to achieve
that outcome within the rules of the game. Consequently, the overall direc­
tion and nature of play in a Hintikka game can be clearly understood.
Even so, a Hintikka game, like a Hamblin game, has a creative aspect.
Players can ask virtually any questions at some points in the game, and play
can therefore be quite wide-ranging. By contrast, in the Lorenzen games,
strategy is dictated by the procedures of classical logic so that these rules
regiment the discussion.
We recall that in a Hintikka dialogue-game there are two types of
moves. A deductive move consists of a finite number of rules, e.g. rules for
propositional calculus. An interrogative move is a question which must be
given a full, direct answer by the other player. The presupposition of the
question is added to the commitments of the questioner. If a player refuses
to answer, the negation of the presupposition of the question is added to his
store of commitments. The win-loss rule says that a player who deduces his
own thesis from his opponent's commitments wins the game.
In this framework, the notion of an enthymeme seems fairly clear. For
an attacker, a proposition may be assumed to be a premiss of his oppo­
nent's argument only if that proposition is in the opponent's commitment-
set. For a defender of an argument, one may assume a proposition as an
 ENTHYMEMES 141

enthymeme only if it is in one's own commitment-set.

One problem with this approach is that it does not single out the
unique enthymeme. Suppose that the defender enunciates premisses one
and two of the argument below. We also know, let's say, that premisses
three and four are contained in his commitment-set.
All tall men are mortal.
All short men are mortal.
Socrates is a tall man.
Socrates is a short man.
Therefore, Socrates is mortal.
Which premiss is the enthymeme? Three or four? One answer would be to
select the disjunction of three or four (the weakest proposition needed to
deduce the conclusion). Another is to note that in a Hintikka dispute, it
really doesn't matter. One is as good as the other to prove the conclusion.
And that, after all, is the whole point of the game.
This solution to the problem of enthymemes is not bad. But there are
three reservations we should register. First, it seems to go too far in the
direction of (a) from section 1. It is exclusively up to the proponent of the
argument whether or not a proposition is an enthymeme of his argument.
This observation leads to another reservation.
Usually an enthymeme is some proposition that the person to whom an
argument is directed would assume or may be expected to assume. It is not
a proposition that he definitively or explicitly assumes or accepts. But in
Hamblin and Hintikka games, the commitments are public statements.
Whether a proposition is a commitment always admits of a clear yes-or-no
answer. Simply check that player's commitment-set and see if the proposi­
tion in question is there or not. If so, it is a reasonable enthymeme. If not,
it is not.
The Hamblin and Hintikka games presume that whether or not a prop­
osition is a commitment of a player is transparently clear. But the doctrine
of enthymemes is useful precisely when this presumption is not met, i.e.
when all premisses are not clearly stated.
Hintikka games are cumulative in the sense that they never allow
retractions of commitments. But Hamblin games like (H) are non-cumula­
tive, and do allow retractions. Consider again the argument above about
Socrates, tall men and short men. In (H) the defending player could retract
either premiss three or four. Consequently, which of this pair the attacking
142 INFORMAL FALLACIES

player chooses as his enthymeme may make a difference. For the defender
may be less likely to retract one than the other. He may find one "more
plausible" or "more central to his position" than the other.
Here then is our third reservation. The notion of 'reasonable
enthymeme' is clear in the Hintikka game, but only because retraction is
not allowed. In a more realistic setting, where retraction may be possible,
the notion of enthymeme remains elusive. It seems to have to do with the
yet undefined notion of what an arguer or audience would assume, rather
than with what the arguer or audience has in fact assumed or conceded.

3. Veiled Commitment-Sets
The doctrine of enthymemes is strategically useful in argumentation
where it may be unnecessary and even an impediment to state all premisses
needed for deductive closure of a conclusion. You can always come back
and plug the loop-holes later provided they are propositions that your audi­
ence would accept, even if they are not explicitly aware of their acceptance.
Hamblin required that the commitment-store of each player be a set of pub­
lic statements, e.g. a number of statements on a slate, in full view of all par­
ticipants. As a variation on Hamblin's theme, let us suggest a second slate
for each participant, not on public view. Let's start with the extreme case
where no player can see his own dark commitment-set, or that of any other
player.
When we say that this "dark" slate is not known to the players, we do
not intend some psychological interpretation of it as "lurking in the recesses
of the player's mind" or some such thing. We agree fully with Hamblin that
there is no place for this sort of psychologism in logical games of dialectic.
The dark-side commitment-set is simply a set of statements, no more no
less. The only difference between our approach and Hamblin's in this
regard is that the "dark-side" set is not on public view to the players. Mem­
bers of it only become known to the players during play of the game,
according to commitment-rules regulating the transfer of statements from
the dark side to the light side of a player's set of commitments.
The following game, drawing some of its characteristics from the
Hamblin game and some from the Hintikka game, differs from both in sev­
eral ways. The most immediately notable difference is that of the player's
dark-side commitment-stores. Unlike ABV, CBV has why-questions.  is
an immediate consequence of A if and only if  follows by a single applica-
 ENTHYMEMES 143

tion of one rule of the game from A. This notion is further explained in
Hamblin (1971), Mackenzie (1981), and Walton (1984).

The Game CBV

Locution Rules
(i) Statements: Statement-letters, A, B, C, ..., are permissible locu­
tions, and truth-functional compounds of statement-letters.
(ii) Withdrawals: 'No commitment A' is the locution for withdrawal
(retraction) of a statement.
(iii) Questions: The question 'A?' asks: Is it the case that A is true?
(iv) Challenges: The challenge 'Why A?' requests some statement
that can serve as a basis in proof for A.

Commitment Rules
(i) After a player makes a statement, A, it is included in his commit­
ment-store.
(ii) After the withdrawal of A, the statement A is deleted from the
speaker's commitment-store.
(iii) 'Why A?' places A in the hearer's commitment-store unless it is
already there or unless the hearer immediately retracts his com­
mitment to A.
(iv) Every statement that is shown by the speaker to be an immediate
consequence of statements that are commitments of the hearer
then becomes a commitment of the hearer's and is included in his
commitment-store.
(v) No commitment may be withdrawn by the hearer that is shown
by the speaker to be an immediate consequence of statements
that are previous commitments of the hearer.
(vi) If a player states 'No commitment A' and A is on the dark side of
his commitment-store, then A is immediately transferred into the
light side of his commitment-store.

Dialogue Rules
(i) Each speaker takes his turn to move by advancing one locution
at each turn. A no-commitment locution, however, may accom-
144 INFORMAL FALLACIES

an a why-locution as one turn.

(ii) A question 'A?' must be followed by (i) a statement 'A', (ii) a
statement 'Not-A', or (iii) 'No commitment A'.
(iii) 'Why A?' must be followed by (i) 'No commitment A' or (ii)
some statement 'B', where A is a consequence of B.

Strategic Rules
(i) Both players agree in advance that the game will terminate after
some finite number of moves.
(ii) The first player to show that his own thesis is an immediate con­
sequence of a set of commitments of the other player wins the
game.
(iii) If nobody wins as in (ii) by the agreed termination point, the
game is declared a draw.
Clearly the main aspect of CBV that makes it so distinctive as a logical
dialogue-game is the addition of a dark-side commitment set for each
player. How this innovation will affect play in CBV and enable us to model
fallacies and arguments in a more revealing way are matters developed in
Walton (1984). Before looking to the special problems posed by
enthymemes, let us review the basic idea behind CBV once again.
The commitment-store of each player is divided into two sides. First,
there is the usual set of commitments resulting from concessions made dur­
ing the course of the game and containing also the initial commitments of
the player. In addition, the commitment-slate of each player has a "dark-
side" — a set of commitments not known to the player or his opponent. As
each move is made in the game, a proposition may come over from the dark
side to the "light side" of the commitment-store. Prior to such a move, the
players might not be completely ignorant of the possible contents of the
dark side of their own or other players' dark side. In some cases, a player
might have a good idea that a certain proposition or its negation may be in
his own or his opponent's dark side commitment-set.
As the game progresses, more and more propositions tend to come
over from the dark side to the light side if the game is progressing satisfac­
torily. It may be that at the end of a game, the dark side is empty, for one
or both players, and the light side contains a large stock of commitments. In
some cases it may be interesting to start a new game with a new set of dark
side commitments, while preserving the light side commitment-sets that
 ENTHYMEMES 145

each player has collected in the previous game. A tournament, or series of

such games, might build up rich stocks of light side commitments.

4. Strategy and Plausibility

The nature of strategy in CBV and related games can be indicated by
the following sort of situation. The two players, Black and White, each
have already made certain commitments.

White's thesis is IB. Black's thesis to be proven is B. Each looks around for
assumptions needed to yield deductive closure of his own thesis from his
opponent's commitments by the following rules.
S ⊃ Τ, S, therefore Τ: Modus Ponens (MP)
S ⊃ T, therefore
Absorption (Abs.)
S ⊃ T , therefore
S ⊃ Τ, Τ ⊃ U, therefore S ZD Ü: Hypothetical Syllogism (HS)
S ⊃ Τ, IT, therefore IS: Modus Tollens (MT)
S ν Τ, IS, therefore Τ: Disjunctive Syllogism (DS)
Black would win if he could get White to accept ID ⊃ B. White would win
if he could get Black to accept ID ⊃ IB. But these would each be one-step
strategies. Neither would accept these respective commitments if they are
playing the game with even minimal skill. Strategic considerations suggest
looking for a more "distant" premiss that one's opponent is not so likely to
immediately reject. For example, Black could select (A ⊃ C) ⊃ (D ν Β) as
a premiss. Or White could select D ⊃ ](B ^ ). Then each would have a
multi-step win-strategy as given below.
WHITE BLACK
I . B ⊃ A Com A ⊃ B Com
2 . ( A ^ B ) ⊃ C Com ( A ^ B ) ⊃ C Com
3.D Com ID Com
4. D ⊃1(B ^ ) Com ( A ⊃ C) ⊃ (D ν B) Com
5.1(B ^ ) 4 , 3 , MP A ⊃ (  ^ ) l,Abs.
6.  ⊃ (  ^ ) 1, Abs. A ⊃ C 5,2,HS
146 INFORMAL FALLACIES

7.  ⊃  6,2, HS DνB 6,4, MP

8.  ⊃ ( ^ ) 7,Abs.  7,3,DS
9.1 8,5,MT
So the strategy of distancing involves linking single applications of a rule
into longer sequences of applications of rules. Otherwise, a player is
strategically set to reject any assumption he sees will directly imply his
opponent's thesis. For that is the nature of the game.
But distancing is only one form of strategy in CBV. If a player needs a
premiss he thinks might be in the dark-side of his opponent, it would be
good strategy to ask for it, even if it directly implies one's own thesis.
Reason: by Commitment Rule (vi), if that premiss is in your opponent's
dark-side store, you will get it into his light-side store immediately, even if
he replies, 'No commitment.' For example, suppose ID ⊃  were in
White's dark-side store. Then it would be good strategy for Black to ask
White to accept ID ⊃ B, even though Black's thesis is a direct consequence
of ID ⊃  and White's previous commitment, ID.
Hence some notion clearly emerges in CBV of what should constitute
a strategic choice of premiss to leave open as a plausible premiss for one's
opponent to accept. The selection is made partly by the attacking player,
who wants to construct a deductively closed argument for his own purposes.
Yet there are certain strategic constraints on what the defending player will
accept as a concession. He will try to withdraw commitment from premisses
that seem to him to permit the attacker good win-strategies. But the choice
of loop-hole-closing concessions is also controlled by a third factor partly
outside the control of both players, namely their dark-side commitment-
sets.

5. The Problem Resolved

The game CBV thus permits a refined solution to the problem of say­
ing what should fairly constitute an enthymeme in an argument. An
enthymeme is not just a proposition that an arguer happens to be commit­
ted to. It is one he is willing to defend and can be prepared to be committed
to, relative to his defence of his position in the context of the dialogue.
Enthymemes are best considered a question of what our trafficking in argu­
ment will bear.
What then is an enthymeme? Suppose I am arguing in order to con­
vince you of a proposition C, my conclusion to be proved in the game of
 ENTHYMEMES 147

dialectic. I have a set of premisses P1 ..., Pi. that are all commitments of
yours, but there is another premiss P. that would make the argument 'Ρ1 ,
..., P., therefore C' valid if added to the premisses. Then P. is a good
enthymeme just to the extent that P. is a strategically sound loop-hole
closer with respect to my strategy in CVB. This means that P. should be a
dark-side commitment of yours, or is a light-side commitment that you will
accept and that leads to deductive closure of my thesis by some (preferably
not too long or too short) finite applications of the set of rules of the game.
This conception of an enthymeme — put in less dialectical and more
rhetorical terms — describes an enthymematic premiss as one that the audi­
ence seems plausibly disposed to accept and that the arguer needs to get a
valid argument for his conclusion. How can the arguer judge the "plausible
disposition" of his target audience? The answer suggested by the context of
CBV is that the arguer must look to the commitments of that audience,
especially to propositions he thinks likely to be dark-side commitments for
that audience. But over and above those factors, he must look to questions
of strategy. Any proposition that appears clearly strategically inimicable to
the audience's own argument — for they too have a position and a thesis to
be proved in any real dialogue — is not a good candidate for an
enthymematic premiss.
That is one kind of enthymeme, but there is another kind as well. The
second kind of enthymeme occurs where the audience (in dialectical terms,
the other player) has constructed an argument for its thesis, but has left out
a premiss, or some premisses, needed to make the argument valid. Here
the speaker has to look to the audience's position and strategy, just as he
did in the first case, and decide what he should reasonably postulate as
enthymematic premisses.
In realistic terms, this second kind of case fits a context where a critic
has surveyed some written or spoken corpus of argumentation for a conclu­
sion. He works over it, trying to fill in the missing bits needed to make the
argument valid. He asks himself — "What was this arguer trying to say?"
Here, unlike the previous case, the critic is not trying to convince the other
arguer per se, but to look over the other participant's argument, and try to
see what it really amounts to.
This second case is more complicated. For when the critic tries to
reconstruct the other arguer's argument, he has to look at himself as a
member of the target audience for that argument, and decide what he
should be willing to accept as reasonable enthymemes for that argument.
148 INFORMAL FALLACIES

Here, he has to consider both his adversary's over-all position and strategy,
as well as his own. It is not only a question of what the arguer seemed to
want to say, or include as premisses, but also what the target audience
(himself included) would be likely to accept as strategically reasonable or
positionally acceptable. Here, the problem is more complex, because both
arguers' positions and strategies may be involved.
In the first case then, I am trying to convince you, let's say, so I look
for premisses you will accept and take for granted without the need for me
to even state them explicitly as part of my expressed argument to you. In
the second case, I am surveying your argument that has been designed to
convince me. I am trying to figure out which propositions you have not
stated which will make your argument valid. Since your argument was
designed to convince me, I need to ask here what I should or might accept
as reasonable premisses, in addition to what you have accepted or might
accept as reasonable premisses in your argument. This second type of argu­
ment reconstruction involves a kind of "double empathy." I need to con­
sider both positions and strategies to fairly reconstruct the argument.
A practically useful doctrine of enthymemes needs to have it that an
enthymeme is a tacit premiss that your audience will accept. Certainly this
much is guaranteed by defining it as a strategic commitment of CBV. But
more than this is involved. An enthymeme is not any proposition your audi­
ence will accept, but one that fits into your own needs as rhetor to have as
an argument that — when properly filled out — is deductively valid.
However, Aristotle makes a distinction between rhetorical and dialec­
tical enthymemes in Rhetorica (1358a 5). Following this cue, we might call
our first type of enthymeme the rhetorical enthymeme. The second, more
complex type of enthymeme we discussed might be called dialectical. But
both need to be analyzed along the dialectical lines set out by the structure
of CBV or its extensions.
Our conception of enthymeme fills both bills, nicely meeting both con­
ditions (a) and (b) from section 1 without being over-accommodating to
either. Enthymemes can only be both practically and theoretically made
useful and understood in the context of strategy in a game like CBV.
The above analysis of enthymemes, I believe, does justice to the
requirements of van Eemeren and Grootendorst that the missing premiss
be an element of the speaker's commitment that he is obliged to defend.
However, it may do less justice to their requirement that the missing pre­
miss must be an informative statement. I think that is for the reason that
 ENTHYMEMES 149

CBV defines 'commitment' and 'obligation to defend' precisely and

appropriately, but does not serve to define 'exchange of information' very
well. To do that, I think we need to explore Hamblin's dialogue-objective
of "information-orientation" more fully. That could be a future project.
However, we have gone far enough at least to see how games with vei­
led commitment-sets can be very useful in working towards a more
adequate conception of enthymemes. The problem is by no means entirely
solved yet. The next chapter will suggest another factor in filling in missing
premisses in extended discourse related to looking at argumentation as a
sequence of steps in dialogue.
From a practical point of view then, what does our solution to the
problem of enthymemes offer by way of advice to the arguer confronted by
an argument that evidently has some missing premisses? Let us look at the
two kinds of cases — first the rhetorical case and then the dialectical.
First, suppose an arguer wants to convince his audience that Socrates is
mortal, once having advanced the premiss 'All men are mortal' and his
audience has accepted that premiss. Should he consider 'Socrates is a man'
as the appropriate enthymematic premiss? According to our theory, the
problem becomes one of argument strategy. If the audience has acknow­
ledged its commitment to 'Socrates is a man' or that proposition is a dark-
side commitment of that audience, then it may be reasonable for the arguer
to treat that proposition as an enthymematic premiss. If, however, the audi­
ence strongly rejects the conclusion 'Socrates is mortal,' then treating 'Soc­
rates is a man' as an enthymematic premiss is not warranted. The audience,
once aware that these two premisses imply the conclusion they strongly
reject, will simply reject one of the premisses. This being the case, the
arguer should be advised to look around for other propositions more appro­
priate to play the role of enthymematic premisses. The problem of rhetori­
cal enthymemes is thus shown to be an instance of the general problem of
strategy in argument faced by any arguer who wishes to convince a target
audience of his conclusion.
Let's now consider the dialectical case of the critic who isn't sure
whether or not he can fairly add the premiss 'Socrates is a man' to 'All men
are mortal,' given that the argument he wishes to evaluate has as its conclu­
sion 'Socrates is mortal.' His problem may be more complex if the propo­
nent of the argument at issue is not around to accept or reject the proposed
enthymeme as representative of what he wished to say. If he postulates as
a missing premiss a proposition that the arguer has failed to give reasonable
150 INFORMAL FALLACIES

evidence that he would accept as part of his position, the critic has then
committed a straw man fallacy. How is he to avoid this fallacy?
The critic must put himself into the position of the proponent of the
enthymeme, and postulate as a premiss to fill the gap some proposition that
fairly represents that position, or at least does not run counter to it. But
second, he must pick a premiss that, in the view of the proponent, would be
strategically appropriate to convince the target audience towards which the
enthymeme was directed. Hence the critic must ask: does 'Socrates is a
man' fill both these requirements? Only if it meets both conditions should
he reasonably propose it as an enthymematic premiss. If not, he may look
around for some other proposition that fairly meets both criteria.
Once again then, the general problem in the dialectical case for the cri­
tic is just a special case of the strategic problem that confronts any arguer in
a logical dialogue-game. How can I select premisses that the other partici­
pant will concede and that will allow me to prove my conclusion by the log­
ical rules of the game? Although we have now given a general framework
for solving the problem of enthymemes, some particular, practical prob­
lems still remain.
Suppose an arguer is confronted by an enthymeme that he could use to
convince someone of a conclusion. But suppose this case is complicated by
the fact that there are evidently two or more missing premisses needed to
fill the gap. As far as the theory of argument is concerned, could the order
in which these premisses are filled in make any difference? Or is the order
of the premisses in an argument irrelevant to considerations of argument
strategy? In the next section, we will prove that the order of premisses is
both relevant and important.

6. Order of the Premisses

Consider the following problem posed in Walton (1984, p.214). Sup­
pose I am presenting an argument in a dispute to a respondent. My thesis to
be proved is C. Hence the respondent is inclined to reject C. But at the pre­
sent stage of the argument, he has just accepted IB. The problem is: which
premiss should I propose next, A ⊃ B O r A V C ?
 ⊃ 
A v C
IB

 ENTHYMEMES 151

Can the order of presentation of the premisses make a difference to the

respondent's play?
Suppose he is inclined to accept IB, and has in fact conceded IB. Now
there are two possibilities: to propose A ⊃ B first, or to propose A v C
first. Let's take it that A ⊃  is proposed first. Then the respondent is likely
to reject A. 1 This means that when A v C is proposed as the next move, he
will be inclined to reject it. Why? Because he already rejects  (the oppo­
site of his own thesis to be proved). So if he rejects both A and C, he will
reject A v C.
But now let's see what happens if the premisses are proposed the other
way around, i.e. suppose A v C is proposed first. The respondent rejects
C, but that does not mean he will necessarily reject A v .  might think
that A by itself is a plausible proposition. If so, he might accept A v C,
even though he is set to reject C. What will happen next, when  ⊃  is
posed to him? Well, he rejects B, so he is likely to reject A as well, once he
accepts A ⊃ B. But that does not mean he will necessarily reject  ⊃ . If
he accepts the required connection between A and B, he may accept A ⊃
B, even though he rejects B.
It seems then that the order of presentation can make a difference in
strategy. If A ⊃  is presented for acceptance first, then the respondent will
likely reject A v C when it is presented next. But if A v C is presented
first, then there seems to be a better chance that the respondent might be
inclined to accept A ⊃ .
There is one hitch. Even if A v C is presented first, the respondent is
going to reject A, once he accepts the next premiss A ⊃ B. He will do so
because he has already rejected B. But having made this move, he may go
back to the previous premiss that he just accepted, A v C. Now he rejects
both A and C. Hence he is inclined to reject A v C.
The two cases we are considering can be represented by the sequences
below.
1.Accept A ⊃ 
2. Reject A v C
According to this sequence of moves, the respondent rejects A, once he
accepts A ⊃ B. But then, since he also rejects C, he rejects A v C at the
next move. The other case runs as follows.
1. Accept A v C
2. Accept A ⊃ 
3. Reject A v C
152 INFORMAL FALLACIES

The reasoning of the respondent in this case is essentially similar, except

that he has already accepted A v C at the first move. But then, having
accepted A ⊃  at the next move, he finds himself to be in an inconsistent
position, and moves to restore consistency by rejecting A v C at the third
move.
Clearly then, there are significant differences between the two cases.
The second sequence is one step longer, and involves the retraction of a
previous commitment.
However, this difference in play as a result of the order in which the
premisses are presented may not be so clear if the respondent plays accord­
ing to another strategy. Even if A v C is presented first, the respondent
may find strategic grounds for rejecting  ⊃  at his next move.
1. Accept A v C
2. Reject A ⊃ B
If the respondent reasons out the following strategy, then the play above
would be determined. I have already accepted A v C, and yet I must reject
 This means that, in a fashion, I am committed to A. For my only permis­
sible grounds for accepting A v C must reside in an acceptance of A. But
if I accept A ⊃ B, then in virtue of my rejection of B, I would be committed
to the rejection of A. But I can't both accept A and at the same time reject
it. Therefore, if I want to be consistent, I had better not accept A ⊃ B. If
the respondent reacts this way, the two cases seem more nearly similar.
Hence, even if A v C is presented first, the respondent may still reject the
other premiss, A ⊃ B .
Despite this evident possibility however, there still remains a crucial
difference. The respondent must be more farsighted if he is to reason, in
the more oblique manner above, to the rejection of A D B . Hence the dif­
ference between the two cases is still highly significant from a point of view
of the strategy of the presenter of the argument.
It is, in general, better to present A v C first, for only a more farsigh­
ted proponent will recognize that his acceptance of  ⊃ , all right in itself,
may get him in trouble in relation to a proposition he has already previously
accepted.
But what does this difference of "farsightedness" in the two cases come
down to? In both cases, it is possible for the respondent to see, once he has
accepted the one premiss, that he can be shown to be committed to a set of
propositions that is collectively inconsistent. Why is it easier for the respon-
 ENTHYMEMES 153

dent to see this possibility in the one case, and harder to see in the other?
To uncover the basis of the difference, we need to recognize that, in
both cases, the respondent applies the same two rules in the same order.
(Rl) A ⊃ B, IB; therefore 1A.
(R2) 1A, 1; therefore 1(A v C)
But to see how these rules are applied, let us review each order of the pre­
sentation of the premisses.
(CI) A ⊃  (2) A v C
Α ν  A ⊃ 
In case (CI), the application of the rules is a simple two-step affair. (Rl)
applies to the first premiss, then (R2) applies to the second premiss, along
with previous commitments. In case (C2) however, the sequence of infer­
ences is not so straightforward. No rule applies to the first premiss, A v C.
No rule can apply yet, because we cannot infer 1A until we apply the appro­
priate rule to the next premiss, A ⊃ B. But then, in (C2), once he gets to
the question of whether to accept the second premiss, he must apply (Rl)
to that single premiss. That doesn't yield any problem. To see the problem
with accepting A ⊃ B, he must then apply (R2) to what has resulted from
applying (Rl) along with his previous commitment to A v C and 1. In
effect then, the respondent must take two steps at the point in (C2) when he
deliberates on whether to accept the second premiss, A ⊃ B .
Now I think we can state what the significant difference is between the
two cases. It has to do with the way the rules are applied at any move in the
game.
Usually a player will consider whether he should accept a proposition
put to him by scanning over each of the rules and applying each rule to the
proposition, taken along with the other members of his existing commit­
ment-set. If this process turns up no inconsistencies, he may feel free to
accept the proposition queried. If he thinks of it, however, he may make a
second pass. He may take the new commitments generated by this first
pass, and go on to ask whether they generate still more commitments, when
the rules are applied to them along with other commitments. Our case (C2)
illustrates that a second pass may turn up new commitments.
It seems then that we need to make a distinction between an immediate
inference and a secondary or tertiary inference. An immediate inference is
one application of a rule to a set of propositions, at some particular move of
the game. But when that inference yields new commitments, a player may
154 INFORMAL FALLACIES

then scan over his newly enlarged commitment-set once again, seeing
whether any rule will apply to some members of this set and enlarge the set
still further.
Here then is the crucial difference between (CI) and (C2). The case of
(CI) represents two immediate inferences. The case of (C2) represents a
secondary inference. First the respondent applied (Rl) to infer 1A. Now
having 1A as a commitment, he scans over his previous commitments and
turns up both 1 and Α ν . Applying (R2) to 1A and 1 yields the nega­
tion of A v C. It is a secondary inference that turns up the problem.
This difference, I think, yields one reason why in much conversational
argumentation, the order in which an arguer presents his premisses is signif­
icant. Secondary or tertiary inferences require a more "farsighted" search­
ing of inferences from one's commitments. Hence inconsistencies of posi­
tion yielded by immediate inferences, even from several lines taken jointly,
are more naturally and easily detected.
We can highlight our findings on enthymemes in the form of a cautio­
nary remark that many textbooks have taken questionable liberties in
assuming that "missing premisses" may be filled in on the basis of loose
conceptions of "charity," or simply a need to make an argument valid.
Exactly which propositions are to be filled in, and in what order, are highly
sensitive and complex matters of argument strategy, very much relative to
the context of dialogue as a contentious process of negotiation.

7. Multiple Premisses in Complex Arguments

We have now developed a theory of enthymemes that may be applied
to cases where an arguer's strategy is relatively clear, and where his argu­
ment is relatively simple. However, the previous section raised the problem
of the order of the premisses. If more than one missing premiss is needed to
reconstruct an argument, how should one proceed? If you have a good idea
of the strategy of the arguer whose argument you propose to complete by
filling in "missing premisses," your decision can be justified by appeal to
your reconstruction of that strategy. But in real life, you may not have a
very good idea of that arguer's most plausible line of strategy. In that case,
your choices of premisses may be hard to justifiably defend as being fair or
reasonable interpretations of this person's real line of argument.
The fact is that when one approaches an argument in a newspaper col­
umn or some other commonplace source, there may be very little informa­
tion available concerning the arguer's position. Picking an obvious way to
 ENTHYMEMES 155

link the premisses and conclusion by some deductive rule of inference may
produce a premiss that the arguer would not accept at all, if confronted with
it. Clearly then, the best principle is to proceed only on the basis of the
information, given relative to the dialogue, on what the arguer's position
really is.
But there is another problem as well. Many arguments in realistic
dialogues are highly complex. They are made up of various stages or sub-
arguments linked together into a complex sequence. In the previous sec­
tion, we have already noted the distinction between an immediate inference
and a secondary inference. In real life, as an argument evolves, a critic or
listener can get more and more of a sense of its direction, its evolution.
Such arguments can be very long, and there could be many constraints on
what sorts of premisses its originator would, or could reasonably allow, in
filling in possible gaps along the way.
In fact many simple arguments with one conclusion, and one or two
premisses, are little more than promissory notes. They could possibly be
backed up by all sorts of powerful and relevant additional arguments and
premisses. Clearly, in such cases, the danger for the critic of committing a
straw man fallacy looms large.
I conclude that the following general policy is appropriate. If a premiss
needed to make an argument valid is clearly both (1) plausible to the
intended receiver(s) of the argument, and (2) plausible as a commitment of
the sender of the argument, as part of his strategy, then adding that premiss
to the argument, and marking it as such, may be justified in argument
analysis.
However, if there is reasonable doubt on the score of either (1) or (2),
then the premiss should not be added in as an acceptable part of the argu­
ment. However, even if convincing evidence of (1) and (2) is lacking, a pre­
miss needed to make an argument valid can be added in, and marked as an
additional premiss that would make the argument valid, but has not been
certified as a premiss to be equated with the other (given) premisses. What
should be clearly marked is that the added proposition has been inserted by
the critic.
So far so good, but the problem is compounded if there are several
ways to fill in the loophole. The problem then becomes: which is the "best"
missing premiss? Moreover, there may be several loopholes, and different
arrangements or sequences of premisses may possibly be plugged in to firm
up the argument.
156 INFORMAL FALLACIES

For this type of case, special techniques to deal with complex

sequences of argumentation are needed. We turn to these techniques in the
next chapter, and the realistic examples analyzed there will illustrate these
practical problems of working with enthymemes. As these examples will
indicate, it is not only missing premisses that may need to be filled in when
dealing with an extended sequence of real argumentation, there may be
several conclusions along the way, and some of these may have to be filled
in as well.

NOTE

1. Suppose the respondent accepts the premiss  ⊃ . He would then be committed to A ⊃

B. But he has previously committed himself to IB. Hence, by modus tollens — which we pre­
sume is a rule of inference of the game — the respondent must be committed to 1A. Hence he
is likely to reject A. At any rate, he is likely to reject A on the assumption that his play is gov­
erned by "logical" defensive strategy.
 CHAPTER 6: LONGER SEQUENCES OF ARGUMENTATION

We saw in previous chapters that two moves in a dialogue-game can

interlock together and form a circular sequence of argumentation. This
observation raises the question of how to keep track of and evaluate the
longer sequences of argumentation in dialogue or other forms of argumen­
tation that are often necessary for the proper development of an argument.
Sometimes arguments are very long indeed, comprising whole books, or
even collections of books.
John Stuart Mill once suggested that even an obviously deductively
valid argument like 'All men are mortal, Socrates is a man, therefore Soc­
rates is mortal' had to be circular. Why did Mill think so? He reasoned that
we could not possibly be assured that the first premiss is true unless we
already know that the conclusion is true. But such a claim is dubious. Could
we not know that the first premiss is true because of biological laws, known
independently of the mortality of one particular individual, Socrates?
To put it another way, the question is this. Which of the following
sequences of argumentation did Mill have in mind?
Socrates is mortal [plus other premisses]
Therefore, all men are mortal.
Socrates is a man.
Therefore, Socrates is mortal.
[Premisses concerning biological laws pertaining to mortality]
Therefore, all men are mortal.
Socrates is a man.
Therefore, Socrates is mortal.
The first argument-sequence is clearly circular. One premiss is in fact iden­
tical to the conclusion. The second argument-sequence may not be circular
however, provided the first premiss can be argued for without presuming
the truth of the conclusion 'Socrates is mortal' as a premiss.
Thus the solution to Mill's puzzle requires a further analysis of the
dialectical background, the longer sequence of argumentation in which the
original single argument-step is to be embedded. 1
158 INFORMAL FALLACIES

1. Sequences of Argumentation
Sometimes when we approach a raw slab of argumentation for
analysis, we see that it is really several arguments joined together in a long
sequence. Some linking statements may be a premiss in one argument, and
at the same time, the conclusion of another argument. The pattern may be,
for example, that ρ is a premiss for conclusion q, then q is a premiss for yet
another conclusion r. In such a case, q is an intermediate conclusion (like a
lemma in a mathematical proof), and we have in effect an argument for r on
the basis of ρ via q. That is, while it is correct to say that there is an argu­
ment from q to r, it is equally correct to say that there is an argument from
ρ to r. Once the two intermediate steps are conceded, the intermediate con­
clusion can, so to speak, be discarded, like Wittgenstein's ladder.
Hamblin (1970, p.299) calls this idea a "thread" or "development" that
involves intermediate statements belonging to neither premisses nor con­
clusion of one single argument. Hamblin challenges the usual idea of the
logic texts that "a complex argument can always be broken down into sim­
ple steps in such a way that, in any given step, there are one or more pre­
misses, just one conclusion and no intermediate statements" (p.229). The
point is that the word "argument" is properly used to denote the complex of
steps as well as the individual steps themselves. Geach (1976, ch.14) re­
introduces some useful terminology for this phenomenon by distinguishing
between argument schemata and themata, where the latter are chains of
schemata linked together to form a more complex sequence. Various illus­
trations in Geach (1976) show how the distinction works in the practice of
argument analysis, but such procedures have long been utilized by logi­
cians, for example in the use of chains of syllogistic reasoning by the
medievals.
Hamblin (1970, p.229) also points out, suggestively for our purposes in
the sequel, that circular arguments may be quite misrepresented if we treat
them as one-step events.
A diagrammatic method of tracking the stages in sequences of
argumentation in a longer passage of argument was devised by Beardsley
(1950). Arguments are sometimes very lengthy and complex — even, in
some cases, comprising whole books — and it can be very useful to organize
an argument into a pattern of smaller steps in order to gain a grasp of its
overall flow and direction.
According to Beardsley, (1950, p.19), one form of structure is the con­
vergent argument, where several independent premisses support the same
 LONGER SEQUENCES OF ARGUMENTATION 159

conclusion. An example would be as follows. If his coat is wet he's been in

the rain 0 . His coat is wet © . Therefore, he's been in the rain (3). Here,
both premisses go together to imply the conclusion.
A second form of structure is the divergent argument, where the same
premiss supports several conclusions. An example would be as follows.

Using circled numerals to represent the respective propositions of the

above two examples, we can contrast the two types of structures above with
a third type, called by Beardsley (1950, p.19), the serial argument.
Convergent Divergent Serial
Argument Argument Argument

In the serial argument, proposition is a conclusion for yet at the same

time a premiss for
An example from Beardsley (1950, p.24) should serve to illustrate how
more elaborate passages of argumentation can be organized into patterns
that involve all three types of structures above.
Nobody in his right mind (except maybe a few hucksters) can deny that
radio programs, taken as a whole, are in a very sorry state: never have
we heard such depressing offerings as the singing commercial and the
audience-participation program. Obviously radio broadcasters need a
new and better code. Statistics show that most of the daylight time is
taken up with soap operas, which bears out my first point, and incidentally
shows that broadcasters underestimate the average person's intelligence
(if that is possible!).
160 INFORMAL FALLACIES

The diagram for this argument given by Beardsley (p.24) looks like this.

However, as Beardsley shows in this case, getting from the raw argument
passage to the diagram may take some re-organization.
Radio broadcasters underestimate the average person's intelligence (if that
is possible!). For statistics show that most of the daylight time is taken up
with soap operas. Never have we heard such depressing offerings as the
singing commercial and the audience-participation program. Clearly,
nobody in his right mind (except maybe a few of the hucksters) can deny
that radio programs, taken as a whole, are in a very sorry state. Radio
broadcasters certainly need a new and better code.
The reorganization proposed above (p.25) by Beardsley does seem to make
the original more orderly, at no cost in fairness to the original argument. At
any rate, we can certainly see the potential benefits of the process.
This diagrammatic method has now been successfully utilized by sev­
eral more recent texts, including Scriven (1976), Geach (1976), Johnson
and Blair (1977), Carlone et al (1981), and in the new sixth edition (1982)
of Copi's Introduction to Logic. However, the general question of precisely
what we are doing when we use such graphlike techniques of argument
analysis needs to be studied. Let us turn to this question.

2. Graphs of Arguments
One method of constructing a schematic analysis of longer sequences
of valid arguments is given by Shoesmith and Smiley (1980). According to
them, the validity of an argument must be the product of two requirements.
Not only must each individual step be valid, but the steps must be correctly
arranged in a sequence. Their use of graphs permits the orderly study of
such arrangements by allowing a set of premisses to have more than one
conclusion. By their reconstruction of the notion of an argument, there may
 LONGER SEQUENCES OF ARGUMENTATION 161

be different routes of argument, each being a valid proof, from the same
premisses to the same conclusion.
Another method which uses graphs to analyze longer sequences of
arguments, although with a different technique and motivation, is that of
Walton and Batten (1984). In this method, an argument is defined as a set
of propositions with one proposition distinguished as the conclusion, the
others called ρ remisses. The system of generating arguments utlizes a set of
rules, which might for example be rules of some deductive logic like classi­
cal PC. A digraph is a set of points (vertices) and a set of ordered pairs of
points called arrows (or sometimes, arcs). Then every argument has a dig­
raph corresponding to it, produced as follows. Every initial premiss is rep­
resented as a point. Each rule is then systematically applied to each pre­
miss, and if a conclusion follows by the rule, a new point is added and given
the name of the new conclusion. Then each arrow is labelled with a number
corresponding to that rule. For example, suppose that p 0 and p 0 =) ρ1 are
initial premisses and modus ponens is a rule (R 1 ). Then part of the graph is:

All the rules and premisses are finite, and of a finitary character, i.e. rules
like p 0 ⊃ (p 1 ν p 2 ν ... ν p n ) for non-finite n, are not allowed. Thus at
some point, the process must stop and be complete. At that point, the
graph displays all possible valid arguments, relative to the rules, from these
premisses to all possible conclusions.
To illustrate this process of argument analysis, consider the following
example. There are three initial premisses: p 0 ⊃ p1, ρ0, and 1 ⊃ 2. The
only two rules are the ones given below.
162 INFORMAL FALLACIES

The graph below shows the argument analysis that results from applying the
rules to the initial and subsequent premisses.

This form of argument analysis shows all the different ways a conclusion
could be arrived at from a given set of premisses. It is the basis of all models
of argument where the "thread of argument" is significant.
Applying this model of argument to realistic case studies of argumenta­
tion involves other steps of preparation. Next, we will look at a realistic
case study.

3. Case Study: Argument on Sex Education

Consider the following sample argument. I have made this argument
up, for purposes of illustration. But it is certainly a realistic type of case that
embodies a controversial issue, and some familiar patterns of argument.
I would like to address the question of whether sex education in the
schools is serving the purpose of decreasing out-of-wedlock pregnancies,
venereal diseases, and promiscuity. My conclusion is based on the fact that
there was a drastic increase in venereal diseases and pregnancies among
teenagers since the introduction of sex education courses in schools. This
increase could have been caused by the introduction of sex education in
the schools or simply by increasingly liberalized attitudes towards sex con­
veyed to the family through the media and other sources. If the latter how­
ever, promiscuity is going to be a result anyway, and consequently sex edu­
cation in the schools isn't going to do any good. So no matter how you look
at it, I conclude that sex education in the schools is not serving its purpose.
Moreover, as sex education is taught in the schools, a factual
approach is taken rather than a moral stance. This way of teaching the sub­
ject encourages a casual and open attitude towards sex. Consequently, stu­
dents lose their respect towards sexual matters and are led to engage in
promiscuous acts. I conclude then that these courses are not serving their
 LONGER SEQUENCES OF ARGUMENTATION 163

purpose, because if one does not respect sex, a person will tend to act
promiscuously.
In order to analyse this argument, let us first sort out the component
propositions that it is made up from.
p0: Sex education in the schools does not serve the purpose of decreas­
ing out-of-wedlock pregnancies, venereal diseases and promiscuity.
p1: There was a drastic increase in venereal diseases and pregnancies
among teenagers since the introduction of sex education in the
schools.
p2: The increase in pregnancies and venereal diseases was caused by
the introduction of sex education in the schools.
p3: The increase in pregnancies and venereal diseases was caused by
increasingly liberalized attitudes towards sex conveyed to the family
through the media and other sources.
p4: There will be an increase in promiscuity among school-age children.
p5: Sex education is taught in the schools by a factual rather than a
moral approach.
p6: The factual rather than the moral approach of teaching sex in the
schools encourages a casual and open attitude towards sex.
p?: Students lose their respect towards sexual matters.
pg: Students are led to engage in promiscuous acts.
Now let us see if we can engage in an accurate and fair reconstruction of the
argument. The ultimate conclusion argued for is ρ . The arguer starts by
advancing the claim p 1 as a premiss, and states that his conclusion will be
based on p1. He then makes the claim that the increase in venereal diseases
and pregnancies could be due to either of two factors — introduction of sex
education in the schools, or liberalized attitudes conveyed by the media.
Plausibly, the arguer is putting forward a premiss of the form 1 ⊃ (p 2 ν ρ3)
here, given that he has already advanced p1: as a premiss.
Then in his next statement, the arguer claims that if the increased pre­
gnancies and diseases are caused by media attitudes, then promiscuity will
result (p 3 ⊃ p 4 ). And he concludes from this claim that, consequently, sex
education in the schools won't do any good (p 0 ). What is he doing here? He
seems to be making a side-argument for p 0 . But how is he getting from p 3 ⊃
p 4 to p0? The most plausible and appropriate enthymematic premiss to put
in here is p 4 ⊃ p 0 . That is, he is suggesting that if there is an increase in
promiscuity (due to media influence), then sex education will not serve its
164 INFORMAL FALLACIES

purpose anyway. This suggests a certain chain of reasoning. Having

advanced the premisses p 3 ⊃ p 4 and p 4 ⊃ p 0 , it suggests that the arguer is
moving towards the interim conclusion p 3 ⊃ p 0 . He is concluding, in short,
that if there are media influences, then sex education will not serve its pur­
pose anyway.
This may not be the only possible interpretation of the line of argu­
ment. But it does seem to offer plausible premisses. The problem is that it
leaves a gap. So far, the argument looks like this.

So far, the first step is valid, but the second step is not. How can we fill in
the needed steps? The way the rest of the argument continues suggests a
reconstruction that could extend this analysis to fill in the gaps in a plausible
strategy.
The argument continues in the second paragraph. First, it adds two
new premisses, p 5 and p 6 . Using these premisses, it draws two conclusions,
p 7 and p 8 . The obvious suggestion for an enthymematic premiss here seems
to be the conditional, (p 5 ^ 6) ⊃ (p7 ^ p 8 ). But possibly the argument
could be reconstructed even more faithfully by postulating that the arguer is
proposing the following premisses: (p 5 ^ p6) ⊃ p7 and  7 ,⊃p 8 . Here, it
does not matter too much, as either reconstruction is valid, and will serve to
derive the conclusion clearly aimed for by the arguer. The following
sequence represents this interpretation.
 LONGER SEQUENCES OF ARGUMENTATION 165

The ultimate direction of the argument is brought about by the enthymema-

tic connection needed for the arguer's evident strategy, namely, P8 ⊃ P0
Hence we can see that the second paragraph strengthens the line of argu­
ment offered in the first paragraph by introducing a separate line of reason­
ing for the same conclusion. More precisely, the argument plausibly pro­
ceeds from p8 to p 4 , and from there to p0.
But we still have our problem that the argument of the first paragraph
is incomplete. Can it be further filled in to lay out a plausible and yet effec­
tive line of strategy?
One way to proceed is to go back through the first paragraph. What is
the arguer claiming? He does seem to be suggesting that either of the fac
tors of the introduction of sex education in the schools or increased libera
attitudes in the media would, by itself, be sufficient to result in increasec
pregnancies and venereal diseases. And in either event, sex education will
not serve its purpose. In other words, in addition to claiming p 3 ⊃ p 0 as a
premiss, he also seems to be committed to claiming p 2 ⊃ p 0 as part of the
argument. If this approach is justified, we can now show how the argument
of the first paragraph fits together into a coherent strategy.
Let's presume that in the context of dialogue, we may assume the fol-
lowing classically valid rules.
166 INFORMAL FALLACIES

Applying the rules to the premisses yields the following graph of the argu­
ment of the first paragraph.

Each rule needs to be applied once to a subset of premisses in order to

deductively yield the conclusion of the argument, ρ .
When we add the supporting argument of the second paragraph, the
following reconstruction of the over-all argument strategy is yielded. To
derive this reconstruction, we need to add two more enthymematic premis­
ses, p 8 ⊃ p 4 and p 4 ⊃ p 0 . They seem to fit into the arguer's position very
well. They are plausible. And they are certainly needed to make the argu­
ment valid. Hence both these premisses are included in the over-all graph
of the argument represented below.
 LONGER SEQUENCES OF ARGUMENTATION 167

Hence we can see how the Walton-Batten method can be applied to give a
plausible reconstruction of a sequence of natural argumentation. The tricky
part is that certain enthymematic assumptions have to be filled in that may
not be explicitly given as the argument is conversationally presented. We
must always remember that such filling-in is based on our own presump­
tions, and that the ultimate test of the appropriateness of a presumptive
commitment of an arguer is whether or not she will agree to it if queried. If
this test cannot be applied, we must try our best to reconstruct the arguer's
plausible strategy. 2

4. Case Study: Circular Argumentation

Circular arguments sometimes occur when there is a feedback situation
or causal loop among a series of events. A typical example occurs when a
citizens committee gets up a petition to improve the bus services to a sub­
urb. They claim that the bus service is poor, that the suburb is well popu­
lated by city workers who commute, that many signatories are in favor of
improved services, and that therefore City Hall ought to improve the ser­
vice.
City Hall, almost by reflex, retorts with the usual argument for a con­
servative and inexpensive policy of fiscal restraint. They claim that not
many people currently use the existing bus service, and that therefore put­
ting additional vehicles on that line would mean empty buses, costly ser­
vices that would not likely be used.
At this point, if nobody tries the experiment to see whether the new
buses would in fact be used, the argument appears to be a stalemate.
However, the citizens committee might retort as follows. The reason
the service is under-utilized at present is just because it is inadequate. The
buses are crowded. Nobody likes to ride on a crowded bus. Moreover,
there are too few buses, with the result that stoptimes are too widely spaced
and therefore not convenient for many commuters.
The feeling of the citizens committee is that it is a Catch-22 situation.
The very reason City Hall gives for not improving the situation is that the
situation is poor. But it is precisely the poorness of the situation that needs
to be corrected. Where can we go from here? Some of the preceding dispu­
tation can be captured in a little more perspicuous form in the following
dialogue.
City Hall: Why should the bus services to this suburb be improved?
Citizens Committee: Because the bus service is poor. Also, the suburb
168 INFORMAL FALLACIES

is well populated by city workers who commute, and many signatories

are in favor of improved services. Both these things are true, and if
they are true, the bus services to this suburb ought to be improved.
City Hall: Why is the bus service to this suburb so poor? Isn't it
because not enough people take the bus?
Citizens Committee: Yes, in a way it is because not enough people take
the bus. If not enough people take the bus there is no incentive to
improve the services. If there is no incentive to improve the services,
the service remains poor.
City Hall: Perhaps, but why is it that not enough people take the bus?
Citizens Committee: Because the service is so poor. If the service is so
poor, fewer people are inclined to use it. Instead, they take their cars.
As reconstructed by the above dialogue, the argument of the citizens com­
mittee is deductively valid. To see why, let the following variables stand for
each of the main propositions of the argument.
p0: The bus services in this suburb should be improved.
p1: The bus service is poor.
p2: The suburb is well populated by city workers who commute, and
many signatories are in favor of improved services.
p3: Not enough people take the bus.
p4: There is no incentive to improve the services.
At its first move in the dialogue, the citizens committee puts forward the
propositions 1 and p 2 , and also the conditional (1 ^ 2) ⊃ ρ0. In its second
speech, the committee asserts p 3 ⊃ p 4 , p 4 ⊃ ρ1, and may be taken to assert
p 3 . At its third move, the citizens committee puts forward the conditional p 1
⊃ p 4 , and may also be taken to put forward the proposition ρ .
The first question to be raised: what do we mean by the phrase "may
be taken to be put forward"? The phrase means that the proposition in
question should be taken as an enthymematic premiss in the argument. In
light of the previous chapter, we can now see that the reconstruction of
extended argumentation by the method of graphs can be enhanced by
applying the theory of enthymemes to fill in missing links. We need to add
premisses where the context of dialogue justifies it. This means applying
the criteria of the previous chapter.
The enthymematic presumption that the committee wishes to put for­
ward p 1 as an additional premiss in the third speech is clearly justifiable is
clear from that speech, and also from the fact that the committee did previ-
 LONGER SEQUENCES OF ARGUMENTATION 169

ously assert p 1 in its first speech. We follow the cumulative presumption of

CBV here in ruling that commitments are stored once more. And, at any
rate, the commitment has not been retracted here.
The assumption that the committee may be taken to assert p 3 as an
enthymematic premiss in its second speech is less straightforward. It
depends on what might be meant by 'because'. Since there is some reasona­
ble room for doubt, let us mark the argument as conditional on the pre­
sumption that the committee would answer 'Yes' to the query 'Did you
mean to assert p 3 as a premiss?' So construed then, the dialogue can be rep­
resented by the following tableau.

The above replies of the citizens committee can each be shown to be a valid
proof of the proposition queried using only two rules of inference.

But it might be interesting to look at the flow of the dialogue as a whole.

The graph below displays the over-all sequence of argumentation. The
graph displays all the propositions in the sequence of argumentation on the
right side of the dialogue-tableau (initial premisses), and shows each valid
step of reasoning. Each proposition that "stands alone" (that does not have
any arrows going into it) is an initial premiss. The graph below is an exam­
ple of the use of the Walton-Batten method to read off and reconstruct the
over-all sequence or direction of an argument from a question-answer
dialogue.
170 INFORMAL FALLACIES

Each proposition appears at a point of the graph. Where a proposition is

deduced by a rule from another pair of propositions (premisses), the
number of the rule, R1 or R 2 , appears on the arrow between the points. So
the graph shows how each step is made, and the over-all direction of the
flow of argumentation.
In this argument, one aspect of special interest is clearly portrayed on
the graph above. There is a cycle in it, (p 1, p 3 , p 4 ). A circle in a sequence
of argumentation appears as a cycle in its graph. A cycle is a set of points ρ 0 ,
..., p n such that there are arrows from p 0 to p1, ρ1 to p 2 , and so forth to p n ,
and also an arrow from p n to p0.
How then are we to evaluate the argument? Does it commit a petitio
principii fallacy? If so, who is the perpetrator?
Our evaluation is that although there is a circle in the citizen commit­
tee's argument, it is not a fallacious (vicious) circle. True, the argumenta­
tion went in a circle, but that is what the citizens committee presumably
meant it to do. They were trying to show to City Hall the feedback situation
implicit in the position that City Hall's rejoinder forced the argument into.
The citizens committee was in effect saying, "Yes admittedly the bus ser­
vice is poor because few people use it. But that's not the whole story. Look
further and you'll see that few people take the service because it is so
poor!" The two factors are connected. You can only change one by chang­
ing them both together, the committee is urging. Thus the committee, by so
arguing, is purposely trying to reveal the inherent circularity of the situa­
tion, not speciously trying to use deceptive circle reasoning to refute City
Hall's argument on inadequate grounds. Hence there is circularity but not
fallaciousness.
The moral of this story is that circularity in argumentation is not always
vicious.3 Thus one must be careful in accusing an opponent in disputation
of the petitio principii fallacy. Mere circularity of argument is not in itself
enough to nail down a charge of having committed ths fallacy.

5. Plausibility Conditions on Arguments

After looking at the previous example of circular argumentation, we
may begin to wonder if circular reasoning is ever fallacious. The question is
raised: precisely what is wrong with arguing in a circle, if or when it is
wrong? The best clue to an answer is in Aristotle's Prior Analytics where
the philosopher wrote that demonstration proceeds from what is certain
and "more prior" as premiss. Aristotle's idea is that in a correct demonstra-
 LONGER SEQUENCES OF ARGUMENTATION 171

tion, the premisses are prior to the conclusion, in the sense of being more
firmly established.4 Thus in a long sequence of reasoning, at each stage the
premisses must stand in a precedence-relation to the conclusion. This pre­
cedence-relation is characteristically asymmetrical. That is, if the premisses
precede the conclusion, the conclusion never precedes the premisses.
From a point of view of the graph of an argument, this Aristotelian
approach amounts to the following requirement: if you pick any two points
on the graph ρ and q, and there is a series of arrows leading from ρ to q,
then the plausibility of ρ must always be greater than the plausibility of q.
That is, the arrows will always take you from the greater to the lower
plausibilities, and never the other way around.
One of the logical consequences of this precedence approach is that
circles can never appear in a graph meeting the Aristotelian requirement of
over-all precedence just stated. The proof of this fact lies in the observation
that if the graph went by a sequence of arrows from ρ to q and thence back
to p, it would follow both that ρ precedes q and that q precedes p. This sit­
uation however is impossible! For the precedence-relation is asymmetrical.
If one proposition precedes another, then the second can never precede the
first. Thus the Aristotelian approach always bans circles.
Sometimes the context of argument indicates that this Aristotelian
requirement of precedence is meant to be met by an adequate sequence of
argumentation. Consider the following dialogue.
Black: How can you determine that juvenile delinquency is wide­
spread in the U.S.S.R.?
White: Well, we know that there has been loss of social control
recently in the U.S.S.R. because of difficult economic prob­
lems. And if there's loss of social control, there must be delin­
quency on a large scale.
Black: Yes, perhaps, but how can you really be sure that there has
been the extent of loss of social control you speak of? After
all, we don't have much access to data.
White: Well, we know that there has been a breakdown in state and
political organization in the cities. If so, there must be loss of
social control to a wide degree.
Black: I hate to be so persistent, but how on earth can you be sure
that there has been such a breakdown in state and political
organization in the cities.
White: Well there is widespread juvenile delinquency, and where
172 INFORMAL FALLACIES

that is present, there is always breakdown of state and politi­

cal organization in the cities.
Black: Hold on a second! Aren't we going in circles?
Black is certainly right. To see why, first identify the propositions in
White's argument.
p0: Juvenile delinquency is widespread in the U.S.S.R.
ρ : There has been a loss of social control in the U.S.S.R.
ρ : There has been a breakdown of state and political organization in
the cities.
Then the over-all sequence of White's argumentation can be represented by
this graph.

Reflecting on our previous case study of the citizens committee, we might

now ask: what is fallacious about White's argument? Surely delinquency,
loss of social control, and breakdown of state and political organization are
all connected in a feedback network. Is it fallacious to point this out?
The answer is "No," but the problem is that the nature of Black's ques­
tions and White's answers make it pretty clear that White means to base p 0
on p1 by a relation of evidential precedence. Then White goes on to base 1
on the prior premiss p 2 . However, when he continues his argument to the
point where he then bases p 2 on p 0 , he violates the Aristotelian requirement
of precedence. He has already backed up p 0 by appeal to the premiss p 2 (via
p 1 ). But now he wants to turn around and back up p 2 by appeal to p 0 as a
basis. He can't have it both ways!
The example shows us that participants in argument should make it
clear whether or not the condition on plausibility of propositions is the
Aristotelian one if the criticism of petitio principii may arise.
 LONGER SEQUENCES OF ARGUMENTATION 173

However, we have to ask whether the Aristotelian requirement is the

only possible condition on plausibility assignments in sequences of
argumentation. The citizens committee case study suggests that it is not. In
fact, reflection will show that the requirement is not appropriate for many
games of dialogue. Most often in dialogue it would be unfairly restrictive to
demand that one's opponent immediately furnish premisses of a higher
plausibility that validly imply the proposition to be proved. Most often an
arguer needs to be given considerable leeway to find premisses that his
opponent will accept as more plausible than the conclusion to be proved to
that opponent by the accepted rules of valid argument. If so, an arguer
often needs interim premisses that are not now known to be more plausible
than the conclusion they imply. Consequently, a weaker condition than the
Aristotelian requirement is often appropriate to adopt for games of
dialogue. Different possibilities suggest themselves here, but one is espe­
cially appropriate.
The emphasis of Rescher (1976) is to select out the least plausible
proposition as the weakest link in a plausible inference. This suggests the
following sort of condition. Look at each sub-argument individually that
has occurred in the over-all network of argumentation in the graph. In each
one, pick out the least plausible premiss. Then amongst all these "least
plausible premisses" pick the greatest. In other words select the max of all
the min. Then the plausibility of a proposition will be equal to the max of
the min of all the arguments for it.
The interesting thing about this type of condition on plausible infer­
ences in dialogues for our present purposes is that it does allow plausibility
values to be consistently assigned to propositions on a cycle of the graph.
Perhaps then, this condition or one similar to it is the appropriate require­
ment on plausible inference for the dialogue of City Hall and the citizens
committee, where the circle is not vicious.
But a serious problem remains. What condition for assigning plausibil­
ity values could be appropriate to represent the context of dialogue where
the circular argument is reasonably judged to be vicious?
The Aristotelian approach we applied in the case pertaining to juvenile
delinquency in the U.S.S.R. was thought to represent a kind of circle that
is open to criticism because it presupposed the convention that the premis­
ses "preceded" the conclusion. What might this mean? One analysis is that
there is a requirement that every proposition in an argument must be such
that each proposition that precedes it as a premiss must have a greater
174 INFORMAL FALLACIES

plausibility than the proposition in question. We can model this analysis as

follows. Let's say that there can be a set of propositions p, q, r, ... , rep-
resented in a graph where there may be an arc (arrow) from one proposi-
tion to another. Where there are a whole series of arrows from one proposi-
tion, p, leading to another proposition, q, we will say that there is a directed
walk (diwalk) from p to q. Let p ~ q represent the idea that there is a
directed walk from p to q. Then the appropriate condition on the plausibil-
ity values of the propositions in the argument to represent the Aristotelian
framework is the one below.
(C1) (Vp) (Vq) (If P ~ q then plaus(p) > plaus(q))
This condition requires that the plausibility values of the vertices on any
graph must be so ordered that as we go along any directed walk we go from
greater to lesser values.
This condition would effectively ban circles in the sense that you could
not, following (C1), consistently assign plausibility values at all on a graph
where there is a cycle. If you had p ~ q and q ~ p, then p ~ q requires
plaus(p) > plaus( q), which contradicts the assignment required by (C1) of
plaus(q) > plaus(p) to q ~ p.
The theory of Rescher (1976) stipulates that the plausibility of the con-
clusion should be at least as great as the least plausible premiss. The appro-
priate condition would then be this one.
(C2) (Vp)(Vq)(If P ~ q then plaus(q) ::::: plaus(p))
Clearly however (C1) is too strong if taken in conjunction with (C2), for
these two conditions, as formulated above, are inconsistent with each
other!
An underlying problem with (Cl) as a general condition for all argu-
ments is that it may not allow a disputant enough latitude in seeking out
sequences of argument that might eventually lead to more plausible premis-
ses. In argument, (C1) demands more plausible premisses immediately,
rather than giving a participant in argument "room to argue."
However, the basic problem with (C2) is that it is too weak to
adequately cover all reasonable contexts of dialogue. For (C2) freely allows
circles anywhere in the graph of an argument.
If the context of dialogue indicates that the appropriate rule is (C1)
then all circles are forbidden. That approach models the requirement that
no circles are allowed, i.e. that all circles are vicious. If the context of
dialogue indicates that the appropriate rule is some weaker requirement
 LONGER SEQUENCES OF ARGUMENTATION 175

like (C2), then circles are freely allowed. This conception could model the
context of dialogue appropriate for the bus service argument. Here there is
a circle, but it is not vicious.
How we should proceed from here in giving a general analysis of the
petitio principii fallacy is an open problem. The approach of Walton and
Batten (1984) is that certain refined variants of (C2)-type conditions are
more generally appropriate. This approach is based on the perception that
many circles in argument are non-vicious. However, where a circle is
thought to be vicious, variants of (CI) should be required.
One thing we have to keep in mind here is that there can be different
games of dialogue in different contexts of argument. That means one
should often ask the question: what is the purpose of this dialogue? We saw
in the first chapter that one main purpose of argumentative dialogue is for
each participant to prove his thesis from the other's commitments (premis­
ses). And it is in this light that the idea of "begging the question" comes to
make sense. If I try to prove my conclusion (thesis) using premisses that
you would not be plausibly committed to — say, because they are identical
with my own conclusion or directly imply it, by the rules — then I have
"begged for" the question. But is such a begging fallacious, or somehow a
vicious petitio principii? The argument of Walton (1984) is that in many
cases, it may not be. Instead, it may be simply poor strategy in this type of
dialogue. And poor strategy in argument may not necessarily be the same
thing as committing a fallacy in argument.
In other words, in many paradigm contexts of argumentative dialogue,
arguing in a circle may not represent a sort of argument that should fairly
be condemned as fallacious. If we look over the graph of the argument, and
see some cycles in it, that may not be a bad thing from the point of view of
the critic, or from the point of view of the one to whom the argument is
addressed.
To show a circular argument is fallacious, the additional step should be
taken of showing that a (CI) type of condition represents a reasonable
requirement for that particular game of dialogue. This might occur where a
certain set of propositions are set aside by the participants as "established
facts" or "evidence" and the conclusion to be proven (which is in doubt) is
to be proven from those prior facts, or from premisses equally well estab­
lished. It is sometimes clear that the context (rules of evidence) does set
requirements of this sort in place relative to a certain game of dialogue.
Yet in other contexts, unless some requirements of this sort does con-
176 INFORMAL FALLACIES

stitute part of the objective of dialogue, circularity of argument may be

judged more benignly. Someone who repeats premisses too often, or goes
around in seemingly pointless circles, may be criticized on various grounds.
But they may not be refuted by virtue of having committed a "fallacy" in
the strong sense — meaning that their argument is flatly refuted. A good
many circles in argument may be of this relatively benign sort.

6. The Missing Links

At any rate, we can now see why some circles in games of dialogue are
to be judged vicious (fallacious) and others are not. It all depends on the
requirements for plausible inference that the participants agree upon. Of
course, in many quarrels and debates, the participants make no such previ­
ous agreements or even think about doing so. Small wonder then that they
cannot agree whether an argument is to be fairly judged fallacious by
reason of circularity or not. Hence if such criticisms are to be fairly dealt
with, rules of dialectic must be established.
It is clear however that many an argument in ordinary conversational
quarrels and debates may be woefully incomplete in other respects as well.
Not only are the rules of inference often left unstated, but sometimes pre­
misses are left out and only hinted at or stated unclearly. Often enough it is
even unclear whether some proposition is meant by its proponent to be a
premiss or conclusion. Small wonder circularity often rears its head but no
decision concerning its fallaciousness can be fairly arrived at!
Thus preparatory to constructing a graph-theoretic analysis from a
dialogue, there is much "cleaning up" of the argumentation to be done. If
the argument is down in front of you in black and white, you may not be
able to approach the arguer and ask him whether such and such a proposi­
tion was really meant to be a premiss in the argument. Perhaps the arguer
may even have died some time ago! Such are the limitations of being a cri­
tic.
Of course, if the arguer is present, you can always ask him. That is why
the game of living dialogue is the best possible model of argument and the
fairest. However, in the absence of an available participant, some scrutiny
of his argument can still be possible if the missing steps are filled in so that
one's opponent is always given the benefit of the doubt. In short, the bur­
den of proof is on the critic, not the victim of the criticism or allegation of
fallaciousness. Let us therefore review the use of graphs to apply to realistic
argumentation.
 LONGER SEQUENCES OF ARGUMENTATION 177

The first task of evaluating an extended train of raw argumentation is

to determine a set of premisses and a conclusion for the argument. The sec­
ond task is to determine the set of rules that are allowed in order to obtain
the desired derivation of the conclusion from the premisses. Once the three
elements — premisses, conclusion, rules — are determined, it is deter­
mined whether or not there exists a demonstration (derivation, solution).
However, to be realistic, this is not all there is to it. The way a train of
reasoning is commonly given, various practical problems complicate evalu­
ation.
First, usually an argument includes some intermediate steps between
the premisses and the ultimate conclusion of the argument. This sequence
of steps is most often not a demonstration, but rather a roughly given out­
line of how the demonstration is presumably meant to proceed. The given
sequence is compatible with some possible demonstrations but not with
others, very often. That is, the complicating factor is that most likely more
than one correct demonstration is possible. As far as the traditional con­
cerns of logical analysis have been concerned, this multiplicity of solutions
does not matter. What matters is the existence of a solution — the choice of
one over another is a mere matter of aesthetics. But if we wish to evaluate
for argument circularity it might matter, because it may be that one solution
is circular, and one not. The intermediate steps, which act as guideposts in
determining which solutions are compatible with the arguments, are there­
fore highly significant in evaluating an argument. For our purposes they
cannot be dismissed, as they can be for the traditional, narrower purposes
of logic.
Consequently, a number of things need to be filled on a graph other
than the designated premisses, conclusions, and rules. The second phase of
analysis is the setting out of the intermediate propositions as vertices of the
graph that mark out the stages in the train of reasoning from the premisses
to the conclusion. Next, we need to put in arcs to show which vertices are
linked as presumed steps in the chain of argumentation. When this is done
we may have a solution or not. If not we may have a partial solution. The
partial solution may be compatible with only one solution or with numerous
solutions.
Thus the unregulated nature of common quarrels and debates leaves
many a gap, and methods of argumentation analysis may be only partially
applicable. That is not the fault of the method or of dialectic however. It is
simply a consequence of the rudimentary state of much conversational
178 INFORMAL FALLACIES

argumentation.
Why use graphs in logic? We conclude this chapter with a number of
general reasons why the analysis of argumentation is enhanced by the
reconstruction method of logical graphs.
1.Picking out the premisses and conclusions of arguments is an important
skill in the use of logic. It should be emphasized to students of logic that in
logical evaluation of arguments it is necessary to correctly identify the pre­
miss proposition and the conclusion proposition in order that the putative
argument can subsequently be evaluated as being valid or invalid. The way
classical logic is normally applied to the evaluation of arguments, the pre­
miss propositions and the conclusion proposition are taken as already desig­
nated. This way of proceeding leaves open the question of whether these
sets have been designated correctly. Using graphs will introduce some new
constraints on the selection of these propositions. It will tell us whether
there are argumentational redundancies in the premisses in the sense that it
may be possible to delete some premisses and still have a valid argument. It
will display clearly what the logical alternatives are in this regard. It will
also identify argument redundancies by separating out sub-arguments that
may be mingled in with premisses but that are not necessary for the conclu­
sion we wish to prove. This process of sorting and clarifying premisses is
especially important in connection with the next use of graphic methods.
Quite generally however, the use of graphs enables us to organize a lengthy
argument and see the overall direction it is taking us in, preparatory to
evaluating the various aspects of the argument more finely.
2. In evaluating long trains of reasoning in, say, a book-length argument,
or complex chain of reasoning, or any argument, it may be very useful to
establish each step in the reasoning and then by means of a graph put all the
steps together in the correct order to gain a sequential view of how the
argument fits together and where it leads. Conversely, it may be useful to
take a complicated web of argumentation, and break it down into indi­
vidual steps to see what its component inferences are. Logical implication
in classical logics is unrestrictedly transitive and does not therefore make
the distinction between an argument "step" and an argument composed of
a sequence of steps. But the method of natural deduction in effect displays
such a distinction by marking off a proof as a sequence of lines each one of
which is a direct consequence of its predecessor by application of a single
rule of inference. Indeed, arrows are often used in natural deduction to
 LONGER SEQUENCES OF ARGUMENTATION 179

keep track of the discharging of assumptions, and the result is a graph-like

display of the proof. Using graphs makes the step-like character of proofs
even more explicit by enabling us to keep track of how we arrived at some
step in a proof from its "distant" predecessors. But in addition, the graph
allows us to tell by inspection what are the possible alternative routes from
one point to another in an argument. This information is particularly criti­
cal in establishing whether a long train of reasoning might contain a circular
argument.
3. A significant reason for using graphs in the analysis of arguments is that
we are enabled to detect "circles" in an argument. The usual approach of
classical logic is not much help in this connection without supplementation
by methods like those we propose, for the argument, 'P therefore P,' is per­
fectly valid despite its blatant circularity — but see Woods and Walton
(1977) concerning the limits of classical logic generally in the analysis of cir­
cularity in arguments. The graph of an argument enables us not only to tell
whether there is a circle, but even to tell whether the proponent of the
argument can break out of the circle in his argument by choosing a "route"
of proof that does not require a circle.
Another advantage of using graphs is that we can not only determine
whether an argument is circular, but we can see how to block the circle by
dropping appropriate premisses or inferences. Thus we are informed by the
graph when a fallacy of petitio principii is committed, but even better, we
are given indication of how the fallacy may be avoided or circumvented in
a specific argument. 5
What a "circle" really amounts to as an identifiable form of logic error
worth studying is a substantive and controversial question. Suffice it to say
here that using graph theory enables us to formalize some significant philo­
sophical intuitions about the logic of circular arguments.
4. One problem we are confronted with in determining premisses in the
analysis of arguments is that of sorting between the case where (i) Ρ1 and P 2
may be taken as two premisses that jointly form, by their conjunction, an
argument for conclusion C, and (ii) P 1 and P 2 are two possible alternative
argument premiss-sets for C, each of which individually constitutes an
adequate argument basis for C. If an argument of the first type is given, it
is a useful task of analysis to be able to determine whether or not it contains
evidentially overdetermined premisses so that it could be broken down into
an argument of the second type with two or more independent sets of pre-
180 INFORMAL FALLACIES

misses, each of which would be individually adequate to establish the con­

clusion. Classical logic by itself is not much help here. If an argument is
valid, you can add as many premisses as you like, and it remains valid (If
the premisses underdetermine the conclusion, then of course the argument
is invalid. So classical logic tells us whether or not the premisses underde­
termine the conclusion. But it does not tell us whether or by how much the
conclusion of the argument is overdetermined by the premisses). It would
be useful in clarifying arguments to be able to break down an overdeter­
mined argument into alternative sub-arguments, if each of the latter consti­
tutes a valid argument independently of its partners. Using graphs allows us
to devise a pedagogically useful procedure for accomplishing this task of
argument analysis.

7. Conclusions on Circular Arguments

The graph of an argument tracks the sequence of reasoning which rep­
resents the evidential backing of a particular conclusion in an argument
relative to a given sequence of dialogue. The graph enables us to identify
circular patterns of argumentation, and then we can go back to the game of
dialogue that provides the context of the argument, and evaluate whether
the circle should be judged vicious or benign. This means that in the theory
of argument, there must be a mapping between the graph of the argument
and the sequence of moves in the corresponding question-answer dialogue.
In fact, however, the pinning down of a given sequence of argumenta­
tion as being clearly and incontrovertibly fallacious is, in many cases, not
that simple. The general procedure for mounting and replying to a criticism
of an argument on the basis of its circularity falls into several distinct stages.
1. The critic must identify the circle in the argument. This means identify­
ing the premisses, conclusions, and so forth, and constructing a graph of the
argument justifiably based on the actual sequence of dialogue recorded.
2. The critic must argue that the circle is vicious rather than benign. He
must show that the circular sequence of argumentation violates some
requirement that can be established as an appropriate rule of dialogue to fit
the context of the argument.
3. Ideally, the critic should confront the arguer who was alleged to have
argued in a circle, and confirm points 1. and 2. Once the critic has
documented and justified these three stages of mounting his allegation, the
burden of proof falls on the arguer criticized to clarify or modify his posi-
 LONGER SEQUENCES OF ARGUMENTATION 181

tion, if he can, in order to rebut the charge. Such a rebuttal can take various
forms.
4. The defender can maintain that his premisses or conclusions have been
wrongly or unfairly identified, or that the sequence of his argument was not
as portrayed by the critic. Using this rebuttal, he can claim that there is no
circle in his argument.
5. The defender can concede that there is a circle, but can maintain that the
circle is benign rather than vicious. Such a defence can take two forms, as
in 6. or 7. below.
6. The defender can argue that although his conclusion may lie on a circle
in his sequence of argumentation, that there is another evidential route by
which his conclusion can be established, thereby making the circle non-wor­
risome. For example, supppose the defender's conclusion is C, and  lies
on a circle in relation to another proposition A in the argument. The defen­
der may claim that there is another evidential basis for C, namely B, that is
not evidentially linked to A, or any other circle, in his argument.

In such a case, the defender admits there is a circle in his argument, but
denies that the circle is vicious. How can he support such a defence? In
essence, he argues that the requirement of evidential priority (greater
plausibility of premisses required by the (CI) condition) can be met. He
concedes that, true, it is logically impossible for A to be more plausible
than  and  to be more plausible than A at the same time. But this circle
of support does not matter, for it remains quite possible for  to be more
plausible than C, provided only that there is no directed walk from  to B.
He claims that there is an evidential route to  — from  — and that is
enough to make the argument non-vicious even if condition (CI) applies to
the dialogue.
7. The defender's other option is to retort that (CI) does not apply to his
argument, given the context of the dialogue. If so, even if there is a circle
in his argument, the defender has committed no fallacy that his critic can or
should complain about. The above sequence of seven stages represents the
182 INFORMAL FALLACIES

best way in which a criticism of petitio principii can be rebutted as benign or

established as vicious. Even where the criticism does stick, and the circle is
serious, however, the defender should always have a chance to modify or
extend his argument to try to cope with or overcome the criticism. Hence
the best and fairest way to conceive the fallacy of petitio principii is as a
dialectical criticism that can be brought forward and defended against in
reasonable dialogue.
This dialectical account of the criticism of petitio fits in with the charac­
terization of the fallacy previously set out in Woods and Walton (1975).
There, we sketched out two basic conceptions of the fallacy.
According to the equivalence conception, an argument is said to be cir­
cular if one premiss, or part of a premiss is either equivalent, or perceived
to be so closely equivalent to the conclusion, that there is no advance. Why
would there be "no advance" in such a case? In light of the present analysis,
there could be no advance if one premiss is identical to the conclusion, for
condition (CI) requires that every premiss be more plausible than the con­
clusion.
According to the dependency conception, an argument is said to be cir­
cular if some premiss depends on the conclusion, or cannot be established
other than by presuming the conclusion as prior evidence for the conclu­
sion. Once again, one might well wonder precisely why such a dependent
argument must be fallacious? In our article (1975) Woods and I puzzled
about this problem and attempted to explore some explanations. But in
light of the analysis of the foregoing chapter, the answer has become much
clearer. If an argument is circular and there is no path of argument to the
conclusion from any proposition not on a circle also including the conclu­
sion, then the requirement that the premisses be more plausible than the
conclusion could never be met. Why not? The answer is that 'A is more
plausible than B' is an asymmetrical relation. If A and  are both on the
same circle in an argument, then if A is prior to ,  cannot be prior to A.
You can have evidential priority one way or the other, but not both ways at
the same time. To make A a premiss for  in a plausible argument with
condition (CI) in place, you must have some other evidential route to  not
including A, if there is a line of argument to A that includes  as a pre­
miss.
The long and the short of our analysis of the petitio principii is that you
have to be very careful in mounting this criticism that you get the argument
right that you intend to criticize. What we have seen is that, in real life,
 LONGER SEQUENCES OF ARGUMENTATION 183

there may be many gaps and uncertainties in fairly reconstructing another

arguer's sequence of argument. A good criticism of petitio therefore
depends heavily on the resources of argument analysis built up in the
foregoing six chapters.
In the fifth section of the last chapter, we will apply these lessons to
some realistic case studies of criticisms of petitio principii in scientific
research. These case studies will bring out very graphically the care needed
in approaching real problems of evaluating petitio criticisms if practical logic
is to be a genuinely useful and helpful discipline in supporting serious
research and inquiry in specialized areas.

NOTES

1. For a fuller outline of Mill's puzzle, see John Woods and Douglas Walton, 'Petitio and Rel­
evant Many-Premissed Arguments,' Logique et Analyse, 77-78, 1977, 97-110.
2. See Hamblin (1970) for a further discussion of enthymemes.
3. A more complex example with the same moral is given in Walton and Batten (1984).
4. A fuller exposition of Aristotle's doctrines of circular argument is given in John Woods and
Douglas Walton, 'The Petitio'. Aristotle's Five Ways,' Canadian Journal of Philosophy, 12, 1982,
77-100.
5. Additional illustrations of the use of graphs to study circular argumentation are given in
Walton and Batten (1984). The concept of a graph and its use to model arguments are devel­
oped in a more rigorous way in that article.
 CHAPTER 7: FALLACIOUS ARGUMENTS FROM AUTHORITY

One of the most common types of fallacy cited in traditional criticisms

of arguments is the so-called ad verecundiam fallacy, or fallacious appeal to
authority. We are all familiar with famous authorities or celebrities of one
sort or another whose prestige is utilized to back up persuasive appeals and
pleas for support, often in a highly questionable way. The problem with
analyzing this "fallacy" is that despite the notorious shortcomings of many
such appeals, sometimes the use of expert advice or testimony appears to
be quite a legitimate form of reasoning if you have no better evidence on
which to formulate policy or make a decision. So how can we tell when such
an appeal is legitimate or when it becomes fallacious?
The first thing to notice is that there are different kinds of authority.
Judicial or legislative authority to enact or enforce a policy is quite different
from the authority of scientific expertise to give advice on a question in an
area of natural science. To make the discussion more manageable, we will
concentrate in this chapter on the authority of "factual" expertise in a field
of knowledge, and say no more about judicial or other forms of authority.
Such appeals to expert judgement are indeed deserving of the label
"fallacy" insofar as they are open to notorious abuses and pitfalls. First, let
us look at some of the shortcomings that must be avoided.

1. How Appeals to Authority Can Go Wrong

There are five basic conditions that must be met for an appeal to the
authority of an expert to avoid fallacy.1 Each condition represents one par­
ticular way in which such an appeal can fail to be adequate. The first condi­
tion stipulates that the source authority must be interpreted correctly. This
condition, although it sounds trivial, is nonetheless a requirement that is in
practice quite difficult to meet. The problem is that experts often speak in
the technical language of a particular discipline, and therefore it is in prac­
tice quite difficult for an expert to communicate with someone who is a
layman in that field. It is also characteristically very difficult for two experts
in different fields to communicate meaningfully. Therefore, in quoting the
186 INFORMAL FALLACIES

sayso of an expert it is quite important that the exact words in which the
expert gave testimony be used — or at least as close as possible an approx­
imation be given — so that the expert is not misquoted, and so that subtle
changes in wording may be rendered without a misleading effect.
It is also notorious that experts will attach a number of conditions to
their pronouncements. In many instances an expert will say, not that such
and such will happen unconditionally, but that if certain assumptions are
made, some outcome may occur. However, very often in appealing to
authorities these conditions are not stated, or even overlooked, and the
result can be a disastrous misinterpretation. Omissions of context, for
example, preceding or following the quotation from an expert may radically
effect the statement.
DeMorgan (1847) noted that there is a common practice of putting the
pronouncements of experts together in a fallacious way. For example
expert A may pronounce conditionally that if ρ then q. Then expert  may
then come along and claim that in his view ρ is true. Further, some third
party  may come along and draw the logical consequence q which follows
deductively from both of these premisses, claiming that that is a reasonable
inference to draw from the pronouncements of these two experts collec­
tively. However, in fact it may be the case that neither expert A nor expert
 agrees that q because neither agrees with the premiss of the other. Thus
as DeMorgan wisely warned, it is a common vice to take one premiss from
the individuals of one party, another from others, and to fix the logical con­
clusion of the two upon the whole party. However, such an inference while
deductively correct, may nevertheless be quite fallacious if the conclusion of
the inference is denied by both parties because they disagree with each
other's premisses. We could call this the fallacy of collective inference in
multiple appeals to authority.
The second general condition is that the authority should actually have
special competence in his area of expertise and not simply some superficial
prestige or popularity. It is actually quite difficult to say what constitutes
special competence in an area and this depends on the particular field of
expertise to which one refers. Obviously these kinds of criteria are not
nicely standardized among different fields. Yet perhaps one should take
into account factors like previous record of predictions, tests that the pur­
ported expert may have undergone, or access to qualifications, degrees or
testimony of other colleagues.
The third condition is that the judgement of the expert must actually
 FALLACIOUS ARGUMENTS FROM AUTHORITY 187

be within that special field of competence. Too often what happens in appe­
als to expertise is that the expert may be an expert in field A and therefore
may have considerable general credibility because of the prestige of this
particular field. Yet this legitimate expert may make a pronouncement in
field B, which is not very closely related to field A. Because of a certain
halo effect from area A, added credibility may be given to this person's
judgement in field  simply because he or she is an expert in some area. 2
However, it may be quite difficult to decide whether area A is closely
enough related to area  for the expert's judgement in the second area to
be accorded much credibility over that of a layperson. Therefore we have to
be very careful to recognize that any appeals to expertise in particular fields
are highly topic-sensitive.
The fourth condition is that direct evidence should be available in prin­
ciple if the expert is queried. We presume that for any appeal to expertise
to be adequate and reasonable, the authority should have based his or her
judgement on relevant and objective evidence within the area of expertise.
Of course with appeals to expertise we ourselves may not have direct access
to this evidence.
In fact generally speaking we only appeal to experts, if in fact, it may
be too expensive or otherwise difficult for us to have direct evidence. That
is why we may legitimately appeal to experts as a secondary source of sub­
jective knowledge when we have to make a decision. However, despite this
subjective aspect of the appeal to expertise, the authority should be able to
give some objective evidence to back up his or her judgement if queried.
That is, we presume that the expert has based his or her judgement on
some objective evidence even though this evidence may not be directly
accessible to us.
If there is a panel of experts in a given area who disagree, and the
judgement of one panel member on a certain question falls well outside the
range of the common consensus, then that expert should be called upon to
present evidence for his maverick judgement. This evidence can then be
made available for the evaluation of the other experts. Thus the group
technology of appeal to expertise suggests that where there is disagreement
or cautionary reservation about the testimony of an expert, an independent
evaluation of the evidence upon which that expert based his proposition
should be made. We conclude then that it is a condition of a successful
appeal to expertise that in principle direct evidence must be available.
The first condition is that a consensus technique may be required for
188 INFORMAL FALLACIES

ruling on disagreements amongst qualified experts. It is a commonplace fact

that experts do disagree and in order to adjudicate on disagreements it is
very useful to have a consensus technique. After mutual inspection of the
grounds of disagreement, experts can recast their judgements with the hope
of converging towards agreement.
There have been methods developed, including the so-called Delphi
Technique, for assisting this process of consensus in a panel of experts as an
adjunct to face to face confrontation and panel discussion. The Delphi
Technnique is a method for the systematic solicitation and collation of
informed judgements on a topic that consists of a set of sequential question­
naires interspersed with summarized information and opinion feedback
derived from earlier responses. The intent of the technique is to prevent
professional status and high position from forcing judgements in certain
psychological directions. The intention then is to ensure that the process of
decision-making reflects a true consensus and not simply the psychological
influence of an opinion leader or powerful charismatic personality.
In a round of the Delphi, each member of the panel is asked to write
responses to questions and then the responses are made available to all the
respondents by means of a letter. Any responses that fall outside a range of
consensus can then be reviewed by subsequent rounds of responses. After
several of these rounds one hopes that there will be a reduction in the
spread of opinion. The Delphi then is an attempt to utilize an essentially
subjective means of information to arrive at a decision where one does not
have direct access to available objective evidence. Reliance upon expertise
is not meant to be a substitute for direct experimental evidence, but it is a
useful adjunct to it that enables one to make an intelligent guess if faced
with a decision that can be based on expert knowledge.

2. Plausible Argument
What type of argument is an appeal to expertise? Hamblin (1970,
p.218) suggests that we might at least begin with the deductively valid form
of argument, where E 1 is an expert: Everything E1 says is true, and E1 says
that p, therefore ρ is true. But if Ε1 says that p, and E2 says that lp, and E 2
is also an expert in the sense that everything she says is true, where are we
landed? At least one of them is not an expert, at least in the sense of being
perfectly veracious in what he or she said. Of course we all know that
experts are not always right. Perhaps then, characterizing an expert as one
who, whatever he or she says, always says the truth, is too unrealistic a
 FALLACIOUS ARGUMENTS FROM AUTHORITY 189

definition.
Salmon (1963, p.64) proposes a more realistic inductive model of the
argument from expertise: the vast majority of statements made by E1 are
true, and ρ is a statement made by E , therefore ρ is true. This proposal still
seems too restrictive to capture many appeals to exertise however. For sup­
pose Ε , who is also such that the vast majority of her statements are true,
makes the statement lp. It must follow that lp is probably true. And con­
sequently, it follows that both ρ is probably true and lp is probably true.
But in the probability calculus the probability of ρ is always calculated as 1
minus the probability of p. This means that if the probability of ρ is high,
then the probability of lp must be low. Once again, we cannot have our
cake and eat it too, and must therefore look to an even weaker model of
argument to capture the type of inference characteristic of appeals to exper­
tise. As we will also see with the ad hominem argument in chapter 8, deduc­
tive and inductive logic does not, all by itself, enable us to deal very effec­
tively with the problem of how to proceed when confronted with inconsis­
tency as a datum.
A theory of plausible inference constructed by Rescher (1976) is espe­
cially designed to help us carry on an orderly process of reasoning in the
face of inconsistent data. The plausibility of a proposition is a weaker
notion than either truth or probability, and it has to do with the burden of
proof of the proposition in an argument. If I do not know in fact whether or
not ρ is true, or even probably true, yet I have to act on p, I may decide that
there is a certain burden of presumption in favour of or against p. One way
of judging this might be if ρ is put forward in the argument by someone
else's testimony or conjecture, and I have no reason to question or reject p.
Another way would be where ρ is put forward by some expert witness or
source. Now to say that ρ is plausible does not mean that ρ should be fully
accepted as true, or as a commitment. Rather, ρ should be provisionally
accepted unless some reason for rejecting it comes along. The key idea of
plausible inference is that a conclusion cannot be less plausible than the
least plausible premiss of a deductively valid argument (least plausible pre­
miss rule).
There are two basic cases where plausibility analysis can be applied. If
the given set of propositions making up our plausible data is consistent, we
can determine the plausibility of some statement ρ (contained in the set of
all the deductive consequences of the original set) as follows. Look over all
the arguments, picking out the least plausible premiss in each case, then of
190 INFORMAL FALLACIES

all these, select out the most plausible of the lot. We have already utilized
this rule in chapter 4 in order to evaluate circular sequences of argumenta­
tion. When the given set of data is inconsistent, we have to pare it down by
rejecting some of the propositions.
Suppose we find that a group of experts on a topic disagree among
themselves. For example, on contentious topics like the question of the
safety of nuclear energy, genuine experts may have severe disagreements.
How can we proceed in the face of such an inconsistency? One method,
already mentioned, would be Rescher's method of picking the maximal
consistent set of propositions that contains the most plausible of the propo­
sitions in the original inconsistent set.
A second method would be the already outlined Delphi technique — a
sequence of questionnaires sent out to the experts to try to locate their
areas of disagreement. Then several more "rounds" of questionnaires are
sent out, to try to get the experts to reach consensus by examining the
reasons behind the dissenting opinions. One hopes by a series of Delphi
rounds to achieve a convergence of opinion while avoiding the interper­
sonal psychodynamics of group face-to-face confrontation with its "band­
wagon affect" and dominance of opinion leaders with strong personalities.
This technique is supposed to be an alternative to more direct dialogue
among the experts involved.
But we must ask whether, in the end, there can be any better method
than simply allowing the experts themselves to argue the problem out by
dialogue and criticism of each other's positions. These other methods may
seem more "objective" but could that be because the method of dialogue
seems too subjective as a reasoned way of resolving a dispute or working
towards the truth? Let us look at some case studies.

3. Where Experts Disagree

The Shroud of Turin is a fourteen foot piece of cloth thought by some
to be the burial shroud of Jesus Christ. Stains on the cloth, apparently
bloodstains, outline the image of a man with braided hair, a beard, and a
haunting, ascetic face. Several researchers have examined the cloth when it
has periodically been displayed in a church in Turin, Italy. A laser physicist
has suggested that nuclear energy could have scorched the cloth to produce
the bodily outline displayed on it. Other techniques of analysis used on the
cloth include x-rays, ultraviolet rays, infrared analysis and microscopic
techniques. These seem to show the absence of any paint that might have
 FALLACIOUS ARGUMENTS FROM AUTHORITY 191

been used to produce the outline. Research on the cloth fibers show that it
dates from the time of Christ. Pollen grains found on it indicate that it could
have come from the Holy Land.
Other scientific studies have been done on the anatomical feasibility of
the nail wounds being found on the wrist, as indicated by the shroud. All of
these studies seem to confirm that the shroud could be genuine. Historians
have reasons to be skeptical however. The shroud first surfaced in the four­
teenth century in France, a notorious period for forgery of religious relics.
Some further details of controversies surrounding the shroud are given in
Woods and Walton (1982, p.92).
Some experts believe that the shroud is the genuine imprint of the
body of a man around the time of Christ, but question whether it can be
said that the imprint is in fact that of Jesus Christ. Others question the
shroud's genuine status as a bodily imprint of some natural sort, and are
inclined to think it must be a painting or artifice of some sort, possibly con­
structed during a period later than the era of the crucifixion.
It remains to be seen what inferences can be drawn from this morass of
scientific evidence and expert disagreement. Part of the problem is that the
various pieces of evidence come from different fields of expertise. Con­
sequently, it is difficult to relate them together in an overall evaluation.
Several carefully documented books have now been written about the
Shroud of Turin, and so far the evidence is fascinating but inconclusive. No
doubt further scientific tests will continue.
What is the best method for evaluating complex controversies among
scientific experts on a question like that of the Shroud? What the various
techniques may be said to "prove" appears to remain highly disputable. Sci­
entific methods are involved, but could hardly be expected to yield the con­
clusion that the Shroud is genuine. It seems the best we can do is to refute
the various possibilities of the Shroud's not being genuine. As Sir Karl Pop­
per might point out, we cannot conclusively confirm the correctness of the
hypothesis that the Shroud is genuine. We can only test this hypothesis by
attempting to refute it. Then we can hold the hypothesis provisionally if no
attempt to refute it is successful. The best we can say is that the genuineness
of the Shroud is consistent with tests of the claim to date.
But these tests are conditional upon the state of the art in the various
fields of expertise to which they belong. Since the controversies involved
fall into many fields, perhaps it is better to say that the genuineness of the
Shroud is a plausible conjecture according to the arguments of some
192 INFORMAL FALLACIES

experts, and a dubious conjecture according to other experts. Where does

this leave us? It certainly should not leave us in the position that no conclu­
sion at all can be reasonably drawn on the Shroud of Turin.
But perhaps there is a danger of the ad ignorantiam fallacy here. Cer­
tainly we cannot conclude that the Shroud is genuine and a representation
of Christ. But if there is no conclusive refutation of this hypothesis either,
can we not say that the most intelligent conclusion is that the issue is still
open to argument? That conclusion is non-vacuous, for it denies the claim
that the hypothesis is conclusively refuted or rejected.
But given these uncertainties, how plausible is the claim that the
Shroud is genuine? Surely this conclusion depends on the outcome of
dialogue among the experts, and cannot be intelligently weighed apart from
that dialogue. But why should we have doubted in the first place that the
plausibility of a claim in dispute should be weighed by questioning and
answering of participants in dialogue. Could it be our worry again that
dialogue is too "subjective" a board of arbitration on matters of scientific
truth?

4. Expertise and Legal Dialogue

Dissonance about throwing scientific truth into the arena of question­
ing by non-experts continues to be a strong factor in allowing dialogue as a
legitimate method of argument, criticism and refutation. For example,
many expert scientists called upon to give "expert testimony" in court are
shocked at the way their knowledge is attacked. For one used to the
cumulative methods of scientific experiment and inquiry may tend to per­
ceive legal questioning and cross-examination as both illogical and
threatening. If science is the objective truth, a feeling the scientist may hold
to as an article of faith, how could it be challenged by legal questioning?
The problem for the courts is to reach a verdict of "guilty" or "not guilty"
by looking at the preponderance of evidence in one direction or the other.
In criminal cases, usually the presumption set is that the defendant is
innocent — the prosecution must overcome that presumption by building a
case strong enough to show "beyond reasonable doubt" that the accused
committed a crime. The method is that of questioning both expert and fac­
tual witnesses.
But in law just as in history, as the Shroud issue showed, the facts are
sometimes elusive. Once an event is in the past, we have to rely on the tes­
timony of witnesses as our source of information. Such sources are "subjec-
 FALLACIOUS ARGUMENTS FROM AUTHORITY 193

tive." Hence the objective of dialogue is to shift the burden of proof rather
than to conclusively verify by observation or experiment. If altering the
burden of proof is a worthwhile objective, then why should we deny that
dialogue has a place in the logic of argument?
Moreover, sometimes the testimony of expert witnesses can be ruin­
ously misleading. In such cases, the only safeguard against the ad verecun-
diam fallacy is thorough cross-examination by competent and reasonable
procedures of dialectic. A famous legal case should suffice to illustrate this
point. 1 The "battle of the experts" is notorious in disputes over whether a
defendant was "insane" at the time he committed harm. For psychiatry is
notoriously open to questioning as a science, especially when it impinges
upon issues of morality and responsibility. But even more well-defined
areas like ballistics can become caught up in knotty disagreements between
experts.
In the famous case of Regina v. Roberts (see Canadian Criminal Cases
34 (1977), 177-183 for the appeal), a man was convicted of the murder of a
woman who lived in the same apartment building. The basis of the convic­
tion was expert testimony that loose human hair found on the body, bed
and nightshirt of the deceased were similar to those of the defendant. In his
appeal, the accused man, through the assistance of a concerned attorney
who had taken up his case, sought to introduce the evidence of two other
experts who used different methods of hair analysis. Eventually the appeal
led to a new trial and the man's acquittal.
At the original trial, no expert evidence had been brought forward to
dispute the reliability of the expert testimony that led to conviction. How­
ever, an attorney took an interest in the case, and felt that new evidence
indicated an appeal. He made the appeal successfully, and the case was
then re-tried. It came out during this trial that the original expert had relied
on scientific techniques of hair analysis that were outdated.
The first expert, a specialist in hair analysis in the forensic sciences,
claimed that by looking at patterns and pigments in hair samples, he could
judge when two hairs were "similar" to each other when examined visually
under a microscope. The accused man was convicted, almost exclusively on
the evidence that his hair samples matched those found at the scene of the
crime.
The first expert had been a professor of pharmacology and organic
chemistry. At the appeal and later at the re-trial, his testimony was chal­
lenged by an expert in the field of biochemistry. He agreed that there was a
194 INFORMAL FALLACIES

visual similarity in the hair samples as claimed by the first expert, but
claimed that the method used — that of comparative microscopic analysis
— is subject to a large margin of error in the case of hair like the accused
man's. He also claimed that according to a much more reliable procedure of
testing — neutron activation analysis — the hair samples could possibly
have come from two different persons.
A third expert also brought in to testify was a professor of nuclear
physics and radiochemistry and had done research in radiochemical
techniques for many years. This man originally developed the neutron acti­
vation technique used by the second expert. The third expert ran these tests
on the same hair samples and concluded that it was very unlikely that they
came from the defendant's head.
Based on the new evidence, the accused man was eventually found
"not guilty" and released. In the meantime however, he had spent several
years in prison. Upon release he stated that he was not bitter about his
unfortunate experience, but he would be much more careful another time
in trying to be of assistance to someone in distress. He had not realized
what a compromising situation he had put himself in by trying to render
assistance to the murdered woman.
What went wrong here? In the original trial, the defending attorney
failed to seek out opinions of other experts, even though the judge clearly
attached decisive importance to the testimony of the first expert. Legal
decision-making is based on the model of argument embodied in the adver­
sary system of questioning and challenge of opposing views. But it seems
that it is easy to be impressed by the weight of scientific evidence presented
by an expert authority. The method of dialogue cannot close the argument
at that point however. Both sides of the argument must be represented as
fully as possible, and reasonable care must be taken to secure relevant evi­
dence. Here then is a lesson in caution concerning the essentially contestive
nature of plausible inferences from expert testimony. Examination of both
sides of an issue is ultimately the only safeguard against the fallacious ad
verecundiam argument.

5. Dialogue and Expertise

If the appeal to the authority of expertise is a sometimes reasonable yet
often slippery type of argument, how can we manage or evaluate such appe­
als? We have already noted that the type of argument involved is neither
deductive nor inductive. Where can we turn to some model for grasping this
 FALLACIOUS ARGUMENTS FROM AUTHORITY 195

kind of argument?
We all agree that not all of us can be expert in every field of knowl­
edge. Yet we can sometimes concede that somebody else is an expert even
if we ourselves are not. But that is part of the problem. In a courtroom, and
in many other contexts of argument, the expert must make his case and be
judged by non-experts. And these laypersons must try to arrive at a conclu­
sion based partly on the expert's testimony. Yet they do not have access to
the scientific evidence presumably used by the expert as a basis for his
judgement.
It seems then that each of us has a position in argument, his set of com­
mitments that he is prepared to defend. But the expert has a large and very
complex position presumably based on his scientific knowledge and experi­
ence. We do not have direct access to this pool of information, and we can
only extract it from the expert by a process of question and answer (dialec­
tic). Hence the logic of appeals to expertise reduces to the logic of dialogue.
It's just that the expert is a special sort of participant in dialogue, due to his
special skill or knowledge in special circumstances. We can reasonably
defer to his sayso in some cases unless the dialogue yields a reason to think
otherwise.
Here then is where an expert may be challenged in dialogue by another
expert in the same field. By dialogue and argument, we must decide how to
resolve inconsistencies in the joint testimony of panels of experts. If they
disagree, both cannot be right. They may be able to resolve their disagree­
ments amongst themselves. If not, we may have to decide, in dialogue with
them, who has the best argument. This task is not an easy one for dialectic,
and it is full of pitfalls, as we have seen. But when we have to arrive at some
decision, then dialectic can be a more reasonable method than making a
random guess.
Mitroff and Mason (1981, p.19) go even further and argue that more
importance should be laid on the dialectical examination of an expert's
argument than on any initial presumption about his credentials. They cite
the example of the recent debate over the Vietnam War in support of their
thesis (p.19):
On every issue of importance, there can be found — or created — two
experts, X and Y, who represent opposing views of the issue. It is further
assumed that both experts are equally reputable, creditable, etc., and that
therefore the choice between them is not to be made upon a surface or
superficial examination of either their credentials or their policies.
196 INFORMAL FALLACIES

Mitroff and Mason point out that a Hawk and a Dove on the Vietnam issue
could both be using the identical data in their arguments, e.g. the casualty
figures, economic figures on the cost of the war, and so forth. Yet using the
same data, each supports his opposing view of the war. Thus the view gives
meaning to the data, it seems, rather than the other way around. The data
only become pointful in the argument when coupled with the antecedent
position of the arguer. They conclude that dialectic is essential not only
within science itself, but in using scientific evidence to make decisions on
issues of great importance.
Law textbooks in court practice outline strategies for the novice advo­
cate to manage evidence and procedure in a trial. A sample dialogue in
Cohen (1973, p.535) instructs the lawyer on how to question an expert wit­
ness in the following sort of example: a coroner has testified that death was
due to strangulation, but his autopsy report indicates no trauma to the
neck. The lawyer would like to expose the obvious weakness in this tes­
timony, for normally strangulation would have to mean some trauma to the
neck. However, the expert may quite likely be able to resist that line of
attack if questioned. How should the lawyer approach the problem? Cohen
gives the advice: "Lead him with short, direct questions that generally call
for "yes" or "no" answers." (p.535). One can see the advantages here of
yes-no questions in strategy in dialogue over other types of questions that
permit more freedom to the respondent.
The sample dialogue given by Cohen begins (p.535) with the following
sequence of questions.
Q. Doctor Wright, did you file an autopsy report in the County Clerk's
Office?
A. Yes, I did.
Q. And did you certify that report?
A. Yes.
Q. That report was based upon your examination and autopsy?
A. Yes.
Q. Isn't it fair to say, doctor, that you included in your certified report
all significant data and observations made by you?
A. Yes, that's correct.
Q. And you did not take any testimony from other witnesses under
oath?
A. No.
 FALLACIOUS ARGUMENTS FROM AUTHORITY 197

The lawyer then goes on to ask a series of yes-no questions concerning the
condition of the victim's neck, larynx, and related areas, determining that
there was very little evidence of trauma. Then the questioning continues as
follows.
Q. Doctor, isn't it fair to say that you did not find the classic evidence
of strangulation?
A. No.
Q. Doctor, as a matter of fact, just from your observations alone, you
found no evidence of strangulation?
A. I found ....
Q. Doctor, can you answer my question, yes or no?
A. I found no evidence of strangulation except suggillations in the neck
area.
Here we can see that the lawyer has had to do his homework. He has had
to become knowledgeable to some extent about the nature of the injury and
the relevant physiology, in order to ask the right questions. Even after hav­
ing asked all the relevant questions, and found that there were no injuries
that would be evidence of strangulation, he is still ir trouble, however. For
the coroner has found "suggillations in the neck area." The jury could be
swayed by this possible exception. The lawyer's case would be weakened.
How should he proceed from that point? Cohen (p.536) offers the following
sample of continued dialogue.
Q. But, Doctor, they were not as a result of trauma?
A. I cannot tell.
Q. Did you not say in your report that there was no evidence of
trauma?
A. I said that, yes.
Here, the coroner wavers by answering "I cannot tell," in effect undermin­
ing the lawyer's whole strategy for building up his case. But the strategy
advocated by the sample dialogue above shows careful advocacy. The
lawyer has laid the foundations for his argument by assuring that the
coroner is committed to the accuracy and completeness of his own report,
at the beginning of the dialogue. If you look back to the initial segment of
the dialogue, you recall that the doctor agreed that his certified report
included all his significant data and observations. Now when he vacillates
on the question of evidence of trauma, the lawyer comes back to those ini-
198 INFORMAL FALLACIES

tial admissions. The doctor, to be consistent, must answer to the last yes-no
question by agreeing that in his report there was no evidence of trauma.
So the lawyer has effectively led the coroner to the edge of inconsis­
tency of commitments. When the coroner answers at the later stages of the
dialogue that he cannot tell whether the possible evidence of strangulation
he claims to have found are the result of trauma, he goes against the report
which he is committed in the dialogue to standing by. For at the very end of
the dialogue, the lawyer gets the doctor to concede that the report said that
there was no evidence of trauma. In terms of the theory of games of
dialogue, the doctor's position has been shown to be ambivalent. He is
committed to a certain proposition, but at the same time he has claimed
that he has no commitment to that proposition.
So the lawyer's strategy has been highly effective. The coroner's move
to wriggle out of the conclusion that the lawyer's questions were designed
to lead towards has been effectively negated by the lawyer's over-all strat­
egy in dialogue.
What one should especially notice about the strategy in the lawyer's
dialogue is the careful way the questions are ordered. At the beginning of
the dialogue, the lawyer carefully gets the doctor's commitment to the ver­
acity, accuracy, and completeness of his report. Then later, when the line of
questioning leads in a direction towards the conclusion the lawyer wants to
establish, the doctor sees that conclusion coming, and finds a way to throw
the conclusion into question, whether purposefully or not, it does not mat­
ter. However, having laid the foundations of his argument earlier, the
lawyer can revert to those earlier premisses to close off the loophole. One
key thing to observe here is that in dialogue-strategy, the order of the pre­
misses in a question-answer sequence can be very significant.
This type of cross-examination in trial practice shows that an expert
can be successfully questioned in his own area by a layperson if the layper­
son gets access to the relevant facts and uses them in a strategically well-
constructed dialogue. Hence arguments from expertise can be plausible in
some cases. But in some cases they can be successfully challenged or
criticized by dialectical argument.
The best dialectical method to evaluate or challenge the plausible argu­
ment of an expert is to develop a strategy of yes-no questions that may lead
towards inconsistency of commitments or ambivalence in the position of the
expert. Such dialogue can shift the burden of proof to the other side.
 FALLACIOUS ARGUMENTS FROM AUTHORITY 199

6. Conclusions
Ultimately, the justification of the appeal to the sayso of a qualified
expert comes down to certain practical constraints on decision-making in
actual circumstances. It does often happen that a person has to arrive at a
decision on how to act in a particular situation where it is not possible to opt
out of, or delay the decision. For example, if the decision is whether or not
to elect to take a certain medical treatment, even the decision not to take it
may, in effect, represent a significant course of action. In such a case, the
choice betwen treatment and non-treatment must be made, and it could be
that for the proposed treatment to be effective, the decision must be made
soon.
The first practical constraint obtains when a decision must be made
within a definite time. The second practical constraint which may apply is
that the decision-maker may not have access to the relevant practical skills
or special knowledge to arrive at a reasoned conclusion on how to proceed
without relying on the advice of someone else. For example, the decision­
maker deliberating upon proposed medical treatment may not have time to
acquire the needed medical skills or scientific information relevant to his
problem so that he could make the decision on the basis of his own first­
hand scientific evidence.
Given these two practical constraints, what should the decision-maker
do? Flip a coin? A better method might be to go on the advice of suitably
qualified experts or practitioners who do have the relevant knowledge of
the type of question at issue. If their advice is carefully solicited and
evaluated, it could be a better basis of plausible reasoning that might enable
one to arrive at a conclusion, an "educated guess" on the best thing to do.
If the decision-maker takes care to avoid the pitfalls we have previously dis­
cussed, his conclusion reached through dialogue with qualified experts
could provide reasonable grounds for a plausible conclusion. If no better
source of knowledge is available within the practical constraints of the
problem, reasoning from a position of plausible premisses could be a
reasonable way to arrive at a decision under uncertainty.
However, an objection to this form of reasoning under uncertainty can
be posed. Are we not justifying arguing from ignorance (ad ignorantiam) in
arriving at a decision or conclusion on the negative basis that no better
knowledge is available? Indeed it seems that our justification of relying on
the advice of suitably qualified experts as a form of plausible reasoning does
200 INFORMAL FALLACIES

have the essential form of an argumentum ad ignorantiam. Could it be a

reasonable form of argument from ignorance, like the case we cited previ­
ously of concluding to accept the proposition that a gun is loaded if in fact
we do not know whether it is loaded or not?
To make sense of the ad verecundiam argument as a reasonable type of
argument in some circumstances even if it is a form of ad ignorantiam
reasoning, we have to revert to the basic distinction we made in chapter one
between monolectical and dialectical reasoning. Monolectical reasoning is a
relationship on sets of propositions in relation to reproducible evidence. A
proposition may be said to have properties like being necessarily true, well-
confirmed by evidence, and so forth. All of these properties have to do with
our knowledge that the proposition is true. However, in the absence of con­
firming or falsifying knowledge, we may have to make conjectures on
whether a proposition is plausibly true or not. Such conjectures can be
reasonable in relation to whether the proposition has been reasonably
criticized (refuted) or reasonably defended (criticisms rebutted) in reasona­
ble dialogue. Hence the basis for justifying a proposition as being plausible,
or not, is dialectical. Therefore, a proposition can be reasonably justified as
plausible on a dialectical basis, even if we lack the knowledge, on a
monolectical evidentiary basis, of whether it is true or false. Although
plausible argument may be an argument from ignorance — meaning we lack
knowledge or a monolectical basis for affirming or denying it — it can, in
some instances, be justified as a reasonable form of argument from ignorance.
If an appeal to plausibility on the basis of expert sayso survives dialec­
tical tests and criticisms of the types set out in the foregoing chapter, it may
therefore be judged provisionally reasonable to accept as a commitment in
dialogue. That does not mean it may not be overturned by subsequent
dialogue. All plausible commitments are inherently liable to criticism or
refutation. They can be a reasonable guide to action if no better (monolec­
tical) evidence is available, but they should be regarded, in Rescher's
words, as "fair-weather friends," subject to questioning and review.
In games of dialogue, the initial plausibility values of the propositions
are given. This models the idea that the game modifies these initial values
as strategies are played out, according to the rules of the game. In other
words, the game is played relative to horizon of certain given "accepted
commitments" of the players. This arrangement is very much like realistic
argumentation, where each player approaches an argument with anteced­
ent convictions that may be held with firmer or weaker degrees of convic-
 FALLACIOUS ARGUMENTS FROM AUTHORITY 201

tion. The function of dialectial argument is for the other participant in the
argument to alter the firmness or weakness of those convictions in order to
work towards the proof of his thesis.
When a proposition is brought forward upon the advice of someone
thought by a participant in argument to be a suitably qualified expert, that
proposition may be accorded a certain degree of plausibility by the partici­
pant. But another participant in dialogue can challenge or question the
basis of the proposition, by the means we have examined in this chapter,
and thereby alter the plausibility of it for the other player. The dialectical
study of the ad verecundiam takes this questioning as its subject-matter.

NOTE

1. This case is more fully presented and analyzed in Walton (1984).

 CHAPTER 8: VARIOUS FALLACIES

We have seen that dialectic can incorporate different logics as stan­

dards of correct proof, depending on the nature of the dispute and the con­
ventions adopted by the participants. Classical deductive logic, relatedness
logic and plausible inference each represent different standards of proof.
What about other concepts of valid argument? Can these too be brought
within the scope of games of dialogue?

1. Inductive Fallacies
Two commonly recognized types of fallacies may be called after Sal­
mon (1963) insufficient statistics and biased statistics. These fallacies may
occur when we argue to a conclusion concerning the distribution of proper­
ties amongst a population on the basis of evidence chosen from a sample.
For example, suppose we have a bin full of apples, some red and some
green, but there are far too many to count in the time we have available.
However, we need to know approximately how many of this population of
apples are green as opposed to red. There is a way. On the assumption that
the apples are mixed up in a bin in such a way that the proportion of green
to red is approximately the same throughout, we could choose a random
sample. That is we could pick out a basket of apples selected at random,
and then count the proportion of red apples to green in the sample.
The problem is that there are a number of ways in which this selection
of a sample could go wrong. If we picked only two apples for the sample,
our basis for extrapolating the proportion from the sample to the whole
population would be meaningless. This would be a case of insufficient
statistics. If we picked a sample from a section of the bin we saw to be
mainly made up of red apples, our sample would be misleading. This would
be a case of biased statistics.
To be reasonable, the sampling procedure must ensure that the sample
selected is representative of the whole population in the relevant respect. If
the population is homogeneous, a random sample may be taken, i.e. one
where each item in the population has an equal chance to be in the sample.
204 INFORMAL FALLACIES

If the population is such that the property in question is found in clusters

throughout that population, some form of allowing fairly for the clusters of
distribution must be taken into account.
The fallacies of inadequate statistics and biased statistics are related
because a sample that is very small is much more likely to be biased than
one that is very large in proportion to a total populaton. The important
thing is that if the sample is very small, there is a very great risk of error,
and if the sample is small enough, any statistical generalization may run
such a large risk of error as to be virtually worthless. But a small sample,
even though it may be almost meaningless from a point of view of reasona­
ble sampling conditions, might still impress an audience that might have a
tendency to exaggerate any suggestion of positive evidence because of lack
of critical reflection about the size of the sample.
It is too easy to wrongly infer that what is true of a large sample must
be true of a smaller sample. But this inference could be highly fallacious.
The example of penny-flipping might help to illustrate two fallacies of this
type. Statistical theory tells us that if a penny is a "fair toss," i.e. heads and
tails are equally probable on each throw, then given a large number of tos­
ses, the chances are reasonably good of getting a proportion of "heads" and
"tails" that are roughly equal. But it does not follow that the chances are
also reasonably good of getting half heads and half tails in a small number
of toses. Second, if it is a "fair toss," we take this to mean that each out­
come in a sequence is independent from its predecessors or successors in
that sequence. That is, just because I got heads on this toss, it does not fol­
low that it is more likely that I will get tails (or for that matter heads) on the
next toss. Each toss is independent of the previous one. But we do tend to
regard such sequences as "self-correcting" — for example, given a run of
"tails" we expect the next toss to be more likely to be "heads." This is a fal­
lacy however, and a common enough one that it has a name, the gambler's
fallacy. The fallacy consists at least in part, of confusing a single member of
a sequence with the whole sequence. That is, as we suggested above, statis­
tics tells us that a sequence of a large number of "heads," H H H H H H H
is a slightly less probable group of outcomes than one that contains a "tail,"
H H H H H H T. But if, instead of thinking of the whole group of out­
comes, we concentrate on just one throw, say the last one in the sequence,
then the chances of getting "heads" are just the same as those of getting
tails, namely .5. As far as one single toss is concerned, we must disregard
what went before, assuming again that the coin represents a "fair toss."
 VARIOUS FALLACIES 205

Why is it so easy to commit fallacies of inadequate and biased statis­

tics? According to Amos Tversky and Daniel Kahneman ('Belief in the
Law of Small Numbers,' Psychological Bulletin, 76, 1971, 105-110) people
have a strong intuitive tendency to view samples of a population as being
more representative of that population than sampling theory would indi­
cate. They report, for example, that when subjects are instructed to predict
the proportion of "heads" in a random sequence of fair tosses, they usually
produce sequences that are far closer to .5 than the laws of chance would
allow for short sequences. That is, people tend to have unrealistically high
expectations about the "fairness" of the coin, especially for short sequences
of tosses. Another tendency that people generally have is to think that short
sequences of tosses are "self-correcting," e.g. if we have a string of "heads"
there should be a trend forthcoming towards "tails." Tversky and Kahne-
man suggest that this phenomenon is the heart of the gambler's fallacy:
"The gambler feels that the fairness of the coin entitles him to expect that
any deviation in one direction will soon be cancelled by a corresponding
deviation in the other. Even the fairest of coins, however, given the limita­
tions of its memory and moral sense, cannot be as fair as the gambler
expects it to be." The fallacy, according to Tversky and Kahneman, consists
in the illegitimate inference that what is true of large numbers must also be
true of small numbers. Thus intuitions can be misleading.

2. Deductive and Inductive Arguments

We might now conclude that we know at least something about what
an argument is. At least we seem to know definitively that there are induc­
tive and deductive arguments and how to identify them. Not quite yet, how­
ever. Suppose Sue advances the argument: Most gorillas are vegetarians,
Marvin is a gorilla, therefore Marvin is a vegetarian. And suppose Bob
counters: this argument is deductively invalid, and therefore a bad argu­
ment. Sue might well reply: it is inductively strong, and therefore a good
argument. How are we to determine whether it is really a deductive or
inductive argument?
The fact (at least we may presume it is a fact) that it is inductively
strong but not deductively valid is no help. We know how it measures up
against the standards of each category of evaluation, but that doesn't tell us
which category it belongs to. We are back to the problem of determining
when something is an argument, even if once we've got one, we can utilize
deductive or inductive methodology to tell us whether it is a correct argu-
206 INFORMAL FALLACIES

ment from each perspective or not.

As before, we might suggest that the argument is deductive (inductive)
if the person who advances it honestly intends it to be deductive (induc­
tive), but this question cannot itself be settled by deductive or inductive
rules. Sometimes people are very confused about what they honestly
intend. Thus it is clear that inductive and deductive arguments cannot be
located and identified by the purely inductive or deductive models of argu­
ment through appeal to the internal structures of these models. And it is
clear that an argument can be inductively strong without being deductively
valid. Actually, without appeal to dialectic, there is even something of a dif­
ficulty with this latter proposal, as it is not clear from the internal
mechanisms of deductive and inductive logic whether we want to say of
deductively valid arguments that they are (a) as inductively strong as an
argument can possibly be, or (b) not inductively strong at all, but rather
deductively valid. The safest thing to say, dialectical considerations aside, is
that we have two standards against which arguments may be evaluated as
good or bad, the deductive and inductive standards.
We can summarize the lessons of the discussion as follows. There are
deductive and inductive arguments, but the internal mechanisms of deduc­
tive and inductive logic cannot tell us how to distinguish between the two
classes. Given any set of statements whatever, these two mechanisms will
tell us that certain relations obtain among the statements such that a given
relation is an instance of a good or bad deductive or inductive argument.
All this tells us however is that given an argument, we can evaluate it from
the perspective of the deductive or inductive model, whichever is appropri­
ate. Thus deductive and inductive logic operate within dialectic as compo­
nents of the larger theory of argument. These logics must always take the
argument as given.

3. Post Hoc Arguments

Another fallacy often classified as inductive is the post hoc, ergo prop­
ter hoc argument (after this, therefore because of this). This fallacy is usu­
ally characterized as the argument from correlation to causation, as when
one argues, 'I washed my car and then it rained — therefore one thing must
have caused the other.' However, we have to be careful here to note that
sometimes a correlation between two events is some evidence of the exis­
tence of a causal relationship. Thus the fallacy, more properly described,
consists in the over-hasty conclusion that one event caused the other simply
 VARIOUS FALLACIES 207

because one followed the other. Sometimes correlation is just coincidence,

rather than evidence of a causal connection.
The basic problem with giving a full analysis of this fallacy is that, at
present, there is no widely accepted or well established account of the struc­
ture of causal reasoning. Thus any account of post hoc argumentation has to
pick its way gingerly around this gap.
The general type of fallacy involved in post hoc can be characterized as
follows. If two events A and  are approximately spatiotemporally coinci­
dent, we may say that A and  are related in the sense that A could possi­
bly be a cause of B. However, if simply on the basis of that premiss, we con­
clude that A actually causes B, we have committed a post hoc fallacy. The
error is a modal fallacy of arguing from possibility to actuality.
Until we know more about the analysis of causal inferences however, it
is very difficult to say much very firmly about the positive nature of post
hoc. Instead, the analysis of Woods and Walton (1977) proceeds by post­
ulating several conditions each representing a way in which the post hoc fal­
lacy may occur by failure of correct causal inference. Woods and Walton
(1977) make no claim that the list is complete. Nevertheless, through it we
get some idea of the sort of failures that are involved.
Any correct sequence of argumentation from evidence of correlation
to a causal conclusion requires five basic types of premisses, as below. Let
Φ, Θ and Ψ be events.
(PI) There is a positive correlation between Φ and Θ.
(P2) It is not the case that Θ causes Φ.
(P3) It is not the case that there is a third factor, Ψ, that causes both
Φ and Θ where Φ does not cause Ψ.
(P4) There are no relevant instances of Φ-and-not-Θ.
(P5) Φ is pragmatically relevant to Θ.
Therefore, Φ causes Θ.
With this type of analysis, it is possible to see the variegated nature of post
hoc fallacies. Different types of errors involve the failures of one or more of
the premisses (PI) to P5).
In fact, it turns out to be possible to distinguish several distinct types of
fallacies amongst the traditional treatments of the post hoc.
1. Concluding that Ψ was followed by Φ just because Ψ temporally
followed Φ, or because some instances of Φ were followed by instances of
Ψ. The problem with this type of inference is that by overlooking (P2) to
208 INFORMAL FALLACIES

(P5), it leaves too many gaps. The correlation between Φ and Ψ may just
be coincidence. According to an example often cited by statistics textbooks,
there is a nearly perfect correlation between the death rate in Hyderbad,
India, from 1911 to 1919, and variations in the membership of the Interna­
tional Association of Machinists during the same period. However, to claim
on that basis that these deaths in India were the cause of the fluctuations in
the membership of the machinists association would be a hasty conclusion.
2. Reversing Cause and Effect. Just because there is a correlation
between Φ and Θ, it may not be clear whether the correlation is due to the
relationship that Φ causes Θ or that Θ causes Φ. For example, there may
be a correlation between wealth and participation in mutual funds. But
which causes which? In the absence of further information, it may be dif­
ficult to arrive at a conclusion. Campbell (1974) relates the following exam­
ple of a market researcher, asked to find out if spending more money on
advertising would increase sales. He selects a random sample of businesses
which shows a strong positive correlation between sales in dollars and
advertising expenditures. What to conclude? Perhaps the businesses with
high sales are also the ones most inclined to spend a lot on advertising. The
possibility is that the causal relationship might be going from sales to adver­
tising, rather than the other way around. Thus our market researcher
should be careful about rushing to a quick conclusion, at the risk of commit­
ting a post hoc fallacy.
3. Concluding that Φ causes Θ when both may be the effects of some
third event Ψ. A metereologist finds that a drop in the price of corn has a
negative relationship with the severity of hay fever cases in the population.
Rather than indicating a causal relationship between the price of corn and
the incidence of hay fever however, a further common causal factor may be
indicated. Perhaps it is the weather. When there is good weather, there
tends to be a bumper corn crop, and corn prices fall. At the same time,
good weather usually means a good ragweed crop, and thereby leads to hay
fever aggravations. Thus it would be misleading to say that the low price of
corn causes severity of hay fever. Rather, both conditions are due to a com­
mon cause — the weather.
4. Overlooking information that may run counter to the trend of a cor­
relation. Suppose we find that there is a correlation between persons who
eat an apple a day and those who visit the doctor less frequently. It would
be quite erroneous to conclude that anyone who eats a bushel of apples
every day would never visit the doctor. The problem here is that a positive
 VARIOUS FALLACIES 209

correlation may hold up to a point, then begin to take on a negative signifi­

cance. This fallacy consists in not taking (P4) into account in causal infer­
ences.
Perhaps we have seen enough to suggest that post hoc fallacies are
indeed significant errors in argumentation, and deserve the serious atten­
tions of applied logic practitioners. It is clear that the efforts of
philosophers of statistics and social sciences to better understand the nature
of causal inferences can be usefully put to work in further research on the
structure of the post hoc fallacy.

4. Slippery Slope
Another type of fallacy having to do with questionable causal infer­
ences is the so-called slippery slope fallacy. A commonly used argument on
the topic of euthanasia is the following.
If we allow non-utilization of aggressive therapies in intensive care
units, we have to allow other waiving of treatments that might shorten life.
For example, we should allow patients to decide not to take chemical treat­
ments for cancer. But if we allow that, it is a short step to allowing patients
to take medications that might have the effect of shortening life, even if the
treatment is not aggressive or painful. And if we allow that, it is just
another short step to allowing patients to alleviate any uncomfortable or
undesirable situation by committing suicide. Once we reach that stage, it
becomes very easy to recommend euthanasia for mentally retarded per­
sons, or anyone who requires inconvenient or costly treatment or support.
Now we all know from the experience of Nazi Germany, it is a short step
from there to elimination of any "socially undesirable," i.e. politically dissi­
dent persons. So once you start in with euthanasia at all, you are on a slip­
pery slope to disaster.
A slippery slope argument of this sort has the following form. Event
A0 is directly related to A 1 , and event A1 is directly related to A 2 , and so
on to A: and through a sequence of steps to some disastrous event A k . So
once you start by engaging in AQ you must inevitably activate some inter­
mediate sequence Ai:, ..., Ak with the end result A k . But the difficulty with
an argument of this form is that the "inevitable sequence" is only as strong
as its weakest link. Moreover, in practice, many intervening steps of the
sequence may be left out, or only roughly sketched in or suggested.
While slippery slope arguments are often partly causal in nature, they
sometimes have to do with the vagueness of terms. The slippery slope argu-
210 INFORMAL FALLACIES

ment above may trade for part of its plausibility on the vagueness of the
term "euthanasia." The first step is one that need not necessarily be
described correctly as "euthanasia." For example if "euthanasia" means
"intentional termination of life (for kindly purposes)" then non-utilization
of an aggressive therapy by a physician in an intensive care unit would not
necessarily be euthanasia, according to many physicians. Presumably, the
doctor's intention may not be to kill the patient, but only to spare needless
suffering by aggressive treatments that would not cure the patient. So the
question is raised: at what point along the sequence of steps does the act in
question become "euthanasia?" Perhaps the participants in the dispute can­
not unanimously decide. If not, the proposer of the slippery slope argument
definitely "euthanasia" as all would agree, and acts that his opponent would
not normally be inclined to agree to as being describable as "euthanasia."
say "Well, that's the insidiousness of it. You can't really say when it gets
started. But once you're into it it's too late to do anything about it in order
to decisively stop it. You're on the slippery slope!" Hence the vagueness of
terms in natural language makes the slippery slope all the more vicious. Its
proponent can claim that there is really no difference between acts that are
definitely "euthanasia" as all would agree, and acts that his opponent would
not normally be inclined to agree to as being describable as "euthanasia".
The slippery slope is similar in some interesting respects to an ancient
type of sophism called the "heap" or the "bald man" argument. The form
exemplified by the example below is sometimes called the sorites argument.
(PI) Every man who is four feet tall is short.
(P2) If you add one tenth of an inch to a short man's height, he still
remains short.
(P3) Every man who is shorter than some short man is short.
Therefore, every man is short.
The third premiss assures us that we include cases of men whose heights are
not multiples of one tenth of an inch. The reason why this argument is cal­
led a sophism is because (i) it is deductively valid, (ii) the premisses are
true, and (iii) the conclusion is false. However, (i), (ii) and (iii) together
constitute a contradiction. For as we remember, a deductively valid argu­
ment is defined as one where it is impossible for the premisses to be true
and the conclusion false. Philosophers do not agree on how to solve the
problem, but most attempted solutions either deny the validity of the argu­
ment or deny the truth of one of the premisses.
 VARIOUS FALLACIES 211

One way to solve this type of problem is to deny the bivalence of clas­
sical logic and instead propose that propositions are true (to a certain
degree) or false (to a certain degree). According to the fuzzy logic of Zadeh
(1975), we should say of the proposition that Bob is short that it is true (or
false) to a certain degree. In other words, instead of propositions taking on
only two values, Τ or F, they could take on a range of values in a so-called
many-valued logic.
By this approach, a solution to the sorites problem is offered by King
(1979). The premiss (P2) is called the inductive step of the argument, mean­
ing that it has a conditional form that can be applied over and over, enough
times to assure validity of the argument. To see how this works, let's estab­
lish a correspondence between the natural numbers and the heights of the
men in the argument. Let 0 be the height of a man four feet tall. Then let 1
represent an addition of 1/10", so that 1 represents the height of a man 4 feet
1/10 inches tall. Then 2 represents another gain of 1/10" and so forth to the n-
th man in the argument. The base step of the argument says that 0 is short.
And the inductive step says that for all the men in the argument, if any man
i is short, then man i + 1 is also short. Hence if you apply this inductive step
over and over again, you will eventually get the conclusion that all men are
short. In other words, the sorites argument is like an indefinitely long chain
of modus ponens arguments.
King's solution proceeds as follows. The conditional that makes up the
modus ponens form of the inductive step is said to be true to a degree. That
is, the propositions in a fuzzy logic have accuracy values instead of truth
values, and these accuracy values can vary from relatively high to relatively
low. Thus the conditional, instead of having an absolute value of "true,"
rather has only a sort of practical legitimacy which gives it a fairly high
value of accuracy. As this conditional is re-applied over and over in each
modus ponens step, the guaranteed minimum value of the conclusion
decreases slightly each time. Thus the premisses of the sorites argument do
not guarantee the conclusion absolutely, because the inductive premiss is
less than absolutely true. In other words, as you keep applying the induc­
tive step over and over, it tends to guarantee to a less extent each time that
the next man it applies to will accurately be measured as "short." Hence the
sorites argument fails to prove that absolutely every man is short.
If fuzzy logic is the right approach to the sorites and related sophisms
due to vagueness, then it is suggested that the best means for studying the
logic of slippery slope arguments may be fuzzy logic.
212 INFORMAL FALLACIES

5. Equivocation
Another fallacy has to do with the meanings of terms in natural lan­
guage, but instead of being based on vagueness of terms it is based on their
ambiguity. A word is said to be ambiguous if it has more than one meaning.
For example, the word 'discrimination' is ambiguous. Sometimes it means
'unfair practice' and sometimes it means 'good taste.' Ambiguity may in
itself be quite harmless, but sometimes it is the basis of the fallacy of
equivocation.
Equivocation occurs where an arguer puts forward an argument that
has an ambiguous term in one premiss, the same ambiguous term elsewhere
(in another premiss or the conclusion, let's say), and where the ambiguity
leaves the recipient of the argument enough scope to go wrong. How can an
equivocal argument go wrong? Let's take an example.
(PI) Marcia showed discrimination in acquiring her lovely new dress.
(P2) Discrimination is illegal in this state.
Therefore, Marcia acquired her new dress illegally.
What appears to have gone wrong with this argument is that 'discrimina­
tion' is most naturally taken to have one of its meanings in one premiss, and
the other in the remaining premiss. But what is wrong with that?
Essentially what is wrong is that the argument is valid and the premis­
ses are true, or so we may presume at any rate, yet the conclusion may be
quite false. A contradiction! How is it to be resolved?
The resolution of this sophism turns on the observation that there are
really four arguments involved, not just one. Because of the ambiguity of
'discrimination,' each of (PI) and (P2) can be taken in two different ways,
one true (or at least highly plausible) and the other false (or at least highly
implausible). Where 'discrimination' means good taste, (PI) comes out
true, but (P2) is false. Where 'discrimination' means unfair practice, the
opposite obtains.
Here is the rub, however. Either way you consistently choose to assign
these meanings to (PI) and (P2), the argument comes out with a false or
implausible premiss. The only way to make both premisses true is to mean
'discrimination' in one sense in one premiss, yet take it in the other way in
the remaining premiss. However, that would make the argument invalid!
Consider,
(PI) Marcia showed good taste in acquiring her lovely new dress.
(P2) Unfair practice is illegal in this state.
 VARIOUS FALLACIES 213

Therefore, Marcia acquired her new dress illegally.

The result of consistency in assignment of meanings is invalidity of the argu­
ment.
The upshot of all this is that we can't get a good argument out of the
four available possibilities — all are either invalid or have false premisses.
In other words, the equivocator is really inviting you to be inconsistent by
taking the premisses in the most plausible way yet also taking the argument
as valid. He is really inviting you to accept more than one argument dis­
guised as one. He is saying "Here. Take one valid one, and one with good
premisses, then you can have both validity and plausible premisses." But
the problem is that it is not the same argument in which these two desirable
properties jointly occur. Hence the fallaciousness of equivocation. We have
merely the appearance, but not the reality of one valid argument with
plausible premisses.
Both equivocation and slippery slope are significant fallacies because
much argumentation takes place in natural language, where vagueness and
ambiguity of terms is not unusual. These fallacies therefore raise questions
about the extent to which dialectic should be formalized. These important
questions about the limits of practical logic demand further study. We
devote chapter 10 to an extended study of the fallacy of equivocation. The
theoretical implications of this particular fallacy have already been made
evident by the chapter devoted to equivocation in Hamblin (1970).

6. Amphiboly
Another traditional fallacy called amphiboly is said to occur when, like
equivocation, an argument is taken two ways. But in amphiboly, the
ambiguity is supposed to be connected with the grammar of a sentence or
phrase, rather than the meaning of a word as in equivocation.
Often however, supposed examples of amphiboly are merely grammat­
ical ambiguities and not arguments at all. A couple of this type of case are:
ROADHOUSE SIGN: Clean and decent dancing every night except
Sunday.
LAUNDROMAT SIGN: Customers are required to remove their
clothes when the machine stops.
These are not even arguments, and consequently cannot be fallacies.
Such a shortage of serious examples of this alleged fallacy leads
Hamblin (1970, p.18) to state that until we get an example of an argument
214 INFORMAL FALLACIES

where the ambiguous verbal construction could be genuinely misleading, it

is hard to take amphiboly seriously as a fallacy. Perhaps the following
example partially fits Hamblin's requirements: Bob said that the shooting
of the hunters is terrible, therefore Bob must be against the killing of game.
The grammatical ambiguity of what Bob said is three-fold. He could have
meant that (1) what the hunters did was a terrible act, a crime, (2) the hun­
ters are bad shots, or (3) the fact that the hunters were shot is terrible. Of
course the conclusion indicates that (1) is meant, and certainly the argu­
ment would not be valid if (2) or (3) were meant.
Clearly this is the sort of argument the textbooks have in mind by
amphiboly. How serious an example it is in Hamblin's sense remains to be
seen. So the question — whether realistic specimens of amphibolous
argumentation can justify the status of amphiboly as a serious fallacy —
remains open to further discussion in chapter 10.

7. Composition and Division

In modern textbooks, the fallacies of composition and division have to
do with incorrect arguments from parts to wholes or vice versa, usually with
physical parts and wholes in mind. To argue that each part of a complex
physical object is light in weight, therefore the whole object is light, would
be clearly incorrect (fallacy of composition). Conversely, to argue that the
whole is heavy, therefore each part is heavy, would be incorrect (fallacy of
division).
The medieval accounts, deriving from Aristotle, often had more subtle
examples in mind. For example, the illustration Aristotle gives in the De
Sophisticis Elenchis (166 a 22 - 166 a 37) is an inference of modal logic.
Aristotle distinguished between the divided sense of 'A man can walk while
sitting,' according to which it is true, and the compounded sense, according
to which it is false. In the divided sense, the sentence says that a sitting man
has the power to walk. In the compounded sense, it says that a man has the
power to walk-while-sitting (presumably, at the same time). We suggested
above that this fallacy is a modal one for it seems to reflect the invalidity of
the modal inference, 'p is possible and q is possible, therefore ρ  q is pos­
sible.' For example, it may be possible that this pencil is red, and possible
that it is not red, but it hardly follows that the statement 'This pencil is both
red and not red' is possible.
Sometimes both modern and medieval examples have to do with the
distributive and collective use of terms, a kind of inference different again
 VARIOUS FALLACIES 215

from the two kinds of examples of composition and division above. One
specimen of composition proposes that it would be fallacious to argue from
the premiss that buses (individually, that is, taken distributively) use more
fuel than cars, to the conclusion that buses (collectively, that is, as a whole
set) use more fuel than cars. However, in our remarks we will comment
only on the most common modern interpretation of these fallacies, which is
concerned with physical parts and wholes.
The basic problem for composition and division is that some but not all
properties are compositionally and divisional hereditary over physical
objects in the following sense. A property is compositionally hereditary with
regard to a collection of parts of an aggregate if, and only if, the whole has
that property if every part has it. For example, if all the parts of a machine
are metal, then the machine is metal. The converse property, going from
the whole to the parts, is that of being divisionally hereditary. The fallacy
occurs when a predicate is mistakenly thought to be compositionally or divi­
sionally hereditary.
Clearly the analysis of this sort of fallacy takes us well beyond classical
logic, or any of the other well-known deductive non-classical logics. In fact,
it requires a theory of collections of physical bodies. There are a number of
such theories, but it is argued in Woods and Walton (1976) that the most
hopeful for this purpose is a theory of aggregates due to Burge (1977).
Aggregates are physical entities in space and time, plural constructions
made up of bodies. Not all parts of an aggregate however are parts that
make up that aggregate. For example, an iron chain is an aggregate, and its
ironness distributes over the parts that make it up, but not over other parts,
e.g. the sub-atomic elements that it contains.
Thus although some of the medieval examples of composition and divi­
sion are amenable to analysis within well-known deductive formal logics
like modal logic, the modern examples clearly demand treatment within
systematic accounts of inference that go beyond the usual range of deduc­
tive logics. While there are resources for this purpose, and future research
on composition and division will undoubtedly benefit from their use, a good
question to pose is how far practical logic wants to go in the direction of
these theories.
 CHAPTER 9: ARGUMENTS AGAINST THE PERSON

We have already encountered the notorious ad hominem fallacy in

chapter one. There, you will recall, we saw that, as traditionally conceived,
this fallacy can take two forms.
The first, the abusive ad hominem, consists in the ploy of avoiding
argument by directing derogatory remarks against the character of one's
opponent in argument. A second form, the circumstantial ad hominem,
involves the allegation that some personal circumstances of one's opponent
are inconsistent with the propositions advanced by the opponent as her
argument. As we saw, the ad hominem argument is often thought incorrect
not least because of its irrelevance.

1. Poisoning the Well

The ad hominem is such a forceful strategic move in argumentation
that it is often quite successful in achieving its purpose of altogether stop­
ping the argument in midstream by undermining the opponent's position.
Sometimes in fact, this refutation is so devastating that further argument is
made pointless or ineffective.
An interesting example of the circumstantial ad hominem argument
occurred in a debate on the topic of abortion in the House of Commons
Debates of Canada (Volume 2, November 30, 1979, p. 1920). In the midst of
a lengthy argument on abortion, the speaker made the following interjec­
tion.
I wish it were possible for men to get really emotionally involved in this
question. It is really impossible for the man, for whom it is impossible to
be in this situation, to really see it from the woman's point of view. That is
why I am concerned that there are not more women in this House avail­
able to speak about this from the woman's point of view.
The speaker is here suggesting that a man can't help being opposed to abor­
tion, or at least taking a particular view on the subject simply because he is
a male person and therefore the topic is in certain important respects, inac­
cessible to his full understanding.
218 INFORMAL FALLACIES

However, it could be pointed out that a counter to this argument is

available to an opponent, namely you can't help being in favor of abortion
or at least taking a certain point of view on the issue because you, the
speaker, are a woman, someone who is in a position to experience
unwanted pregnancy and consequently is unable to resist favoring certain
positions on the issue. This rejoinder creates a stalemate situation in the
dialectic, and in effect stops further discussion, because it implies that nei­
ther side can help taking their particular side or position. The conclusion to
be inferred is that there is simply no point in continuing the argument.
Indeed, an ad hominem move is often so effective in stopping an argu­
ment and undermining an opponent's position that it has been sometimes
called "poisoning the well." This term was evidently coined by Cardinal
Newman who protested an allegation that as a Catholic Priest he did not
place the highest value on truth. Cardinal Newman replied that the accusa­
tion created a presumption that made it impossible for him, or any other
Catholic, to state his case effectively or successfully bring forward further
arguments. Cardinal Newman added that such a presumption made it
impossible for him to prove that he did have regard for the truth, and if he
could not prove that, any argument he might henceforth propound would
automatically be open to the same suspicions and doubts. Thus the term
"poisoning the well" is indeed appropriate in connection with the ad
hominem move, because we can see that once the source of the argument is
impugned, any argument that might subsequently come from that source,
no matter how correct it might be in itself, is condemned by the original
rebuttal of the source from which it proceeds. Thus the ad hominem argu­
ment often has the effect of being a conversation-stopper or blocker of
further argument.
One lesson of this is that whenever there is an ad hominem accusation,
the person accused should always be given fair opportunity to reply to the
charge. Very often however, as in the abortion example, the argument is
cut off so effectively that it is very difficult to know how to proceed from
that point.
It is very natural that textbooks have been inclined to think that one of
the failures of the ad hominem argument consists in a failure of relevance.
In the abortion example, the charge is that the opponent happens to be a
man. This fact may be true, and may therefore appear unchallengeable and
stop the argument. However, what may be overlooked is that the issue has
been changed. The true fact advanced is not directly relevant to establish-
 ARGUMENTS AGAINST THE PERSON 219

ing the claim that abortion is morally right, or its opposite claim that abor­
tion is morally wrong. In some way the topic or issue has been forcefully
changed or re-routed so that the arguer really has no option but to attempt
to respond to the ad hominem allegation, and thereby at least temporarily
drop the main line of argument in order to respond to the opponent's chal­
lenge. This disruption of argument is very often a threatening or upsetting
challenge for the arguer to encounter, and it seems that he loses either way.
If he responds to it, he loses track of his original argument. If he does not
respond, he finds himself unable to further advance his argument with
much effectiveness.
Characteristically, an ad hominem circumstantial attack is mounted on
the basis that there seems to be an inconsistency in the position of the
arguer criticized. Whether there truly is an inconsistency to be found in that
arguer's position then becomes subject to clarification and dispute. Because
the alleged inconsistency is characteristically an ostensible one, sometimes
it is hard to resist the interpretation that the attacker is advancing the
alleged inconsistency more in an attempt to discredit his adversary's posi­
tion or ethics, rather than trying to clarify that position. Sometimes, that is,
one suspects that the ad hominem is merely a diversionary skirmish rather
than a serious attack of the opponent's position.
In the following dialogue from Hansard {Canada: House of Commons
Debates, vol. 24, no. 32, 1983, p. 27379- 27380), Mr. Crosbie, a member of
the Conservative Party, attacks the position of the Liberal Party on the sub­
ject of the economy. (I would like to thank Nora Nolette for drawing this
debate to my attention.)
Hon. John C. Crosbie (St. John's West): Madam Speaker, my question is
directed to the Minister of State for Economic Development. In his brief
to the Macdonald Commission yesterday, the Minister stated this:
Real interest rates are high, and this seems in part a reflection of worries
about a possible resurgence of inflation and the large fiscal deficits in North
America.
He said:
— economic forecasts project a relatively weak economic performance, and
a persistent unemployment problem.
How does this square with the statement of the Minister of Finance of his
Government that "We have a recovery here second to none in the world so
far"?
220 INFORMAL FALLACIES

Who is deceiving the Canadian public, or who is correct? Is the Minister

correct when he says, as he did in his brief, that we are having a weak
economic performance? What will he do specifically to overcome that
economic performance because, in all of this flatulent flapdoodle, there is
no one suggestion of what should be done to overcome persistent
unemployment?
Hon. Donald J. Johnston {Minister of State for Economic Development
and Minister of State for Science and Technology)՛. Madam Speaker, the
document to which the Hon. Member makes reference describes the world
environment as expressed by observers. It brings together the conventional
wisdom of which he perhaps is unaware. Essentially, it deals with the chal-
lenges that lie ahead.
I should add that this submission, which comes from me as Minister of
State for Economic Development, is a paper of challenge. It is not sup-
posed to be a paper of policy. The Macdonald Commission invited us to
put these issues before them for consideration. I found their reaction to be
quite positive, and their questions to be quite intelligent.
Hon. John C. Crosbie (St. John's West): Madam Speaker, my supplemen-
tary question is for the Minister who presented this submission on behalf of
the Government to the Liberal leadership commission meeting in St.
John's, Newfoundland, yesterday. I refer the Minister to page 3. The
Minister refers to a time of only moderate economic growth momentum.
He said that we have a lingering inflation problem, large fiscal deficits, and
all of this militates against a major surge in demand. Who is correct about
our situation — the Minister of Finance who says we have a recovery sec-
ond to none in the world, that demand could be moderated; or the Minis-
ter in his brief where he says that everything militates against a major
surge in demand? How can we have more jobs and economic recovery if
there is a major bias against a surge in demand? Who is correct? Who is
deceiving the Canadian people, and how can this be overcome?
An Hon. Member: A change in government.
Hon. Donald J. Johnston (Minister of State for Economic Development
and Minister of State for Science and Technology): I can see no conflict in
anything that has been said, Madam Speaker.
Some Hon. Members: Oh, oh!
Mr. Johnston: We all know, those of us who have studied economic condi-
tions in the world, that major structural changes have taken place. There
have been changes in demand patterns. There have been changes in world
supply patterns.
Mr. Crosbie: Who is telling the truth?
 ARGUMENTS AGAINST THE PERSON 221

Mr. Johnston: And these have had an impact on our economies. We know
that the demands for consumer durable goods will not be as they were in
the 1960s and into the early 1970s. These are things that are known. This
in no way suggests that we do not have a strong economic recovery under
way. In fact, the recovery is surprisingly high. But there are still concerns,
as the Hon. Member knows, surely as finance critic for the Opposition,
about the level of real interest rates in the United States and also here,
although somewhat less, and this has been interpreted as perhaps a fear
among the population that inflation may recur because of the very large
U.S. deficit. Everybody knows these facts. There is nothing new here.
Mr. Nielsen: Who is telling the truth?
Mr. Johnston: The Minister of Finance has dealt with these issues on
many, many occasions, both in speeches and in the House. I see nothing in
this paper that I submitted that is in any way in conflict with any of the
statements he has made.
Mr. Crosbie: Which one of you is the deceiver?
Mr. Crosbie accuses the Liberal Party of inconsistency. A reported state­
ment of the Minister of State for Economic Development says that there is
a possible resurgence of inflation and large fiscal deficits in North America.
This statement seems to conflict with another, attributed to the Minister of
Finance, that claims that we have a recovery here [Canada?] second to
none in the world. Moreover, the first statement adds that weak economic
performance and a persistent unemployment problem are predicted. As
Mr. Crosbie alleges, these statements seem to run counter to each other.
The pointing out of this apparent conflict is legitimate enough, except
that Mr. Crosbie follows it up by posing a question that is unfair: "Who is
deceiving the Canadian public, or who is correct?" For even if the two
ministers have contradicted each other, it hardly follows that one can be
fairly accused of deceiving the public. Moreover, Crosbie goes on to accuse
the Liberal policy on the economy of being "flatulent flapdoodle," and adds
that it is not doing anything to overcome unemployment. Both of these
statements are packed as presuppositions into Crosbie's complex question
asking what the opposing party is going to do to overcome the weak
economic performance. By posing this series of complex and loaded ques­
tions, Crosbie attempts to make it more difficult for his opponents to deal
with the ad hominem attack he has mounted.
Mr. Johnston replies that the paper quoted by Mr. Crosbie is not a
statement of policy, and describes the world environment as expressed by
222 INFORMAL FALLACIES

observers. However, Mr. Crosbie is not satisfied with this answer, and
repeats his ad hominem allegation and accompanying questions.
Finally, Mr. Johnston replies that there is no contradiction. He points
out that even though demand for consumer products has lessened and
therefore demand levels will never be as high as in the sixties and seventies,
this does not suggest that we do not have a recovery underway. According
to Mr. Johnston then, there is no inconsistency. There is a recovery at the
moment, but there has been a general decline over the longer period, and
there will likely be problems with inflation in the future. Specifying the
times of these fluctuations more precisely, then, has resolved the apparent
inconsistency.
Mr. Johnston's reply is what one expects, and seems reasonable.
Economic indicators are highly relative to particular times and conditions.
However, the Conservative critics persist in their aggressive ad hominem
questions, undaunted. Mr. Nielsen repeats Mr. Crosbie's question, "Who is
telling the truth?" and Mr. Crosbie concludes the question period by say­
ing, "Which one of you is the deceiver?" These repetitions suggest that the
ad hominem criticism was meant more as attack than an attempt at clarifica­
tion of the Liberal position on the economy.
In many cases, what is especially objectionable about an ad hominem
attack is not so much that the criticism of inconsistency is in itself unwar­
ranted or unreasonable, but that it has diverted the dialogue away from
issues that are more serious and germane to the dispute.

2. The Sportsman's Rejoinder

We remember that the classic specimen of the ad hominem argument
was the dialogue we called the sportsman s rejoinder. In this case, the
sportsman accused of barbarity in his sacrifice of game for amusement
retorts, "Why do you feed on the flesh of animals?" This argument is an
interesting one because it turns out to be fairly complex to unravel precisely
what is wrong with it.
In some cases an allegation of circumstantial inconsistency may be in
itself a reasonable move in argumentation, or at least not in itself fallacious.
What is wrong in this case is that the sportsman has shifted somewhat in the
propositions he attributes to his critic in posing the allegation of inconsis­
tency. If the critic himself had been an acknowledged hunter, like the
sportsman he criticizes, then there would have been nothing wrong with the
hunter's pointing out the circumstantial inconsistency inherent in the critic's
 ARGUMENTS AGAINST THE PERSON 223

position. It is as if the critic were to say, "Your hunting of game for amuse­
ment is brutal and vicious," and then the sportsman were to turn and say,
"Aha, but what about you? You are a hunter every bit as much as I am.
What applies to me in this regard must therefore apply equally to you. How
can you engage in sports of the field and yet condemn me for doing the very
same thing. Is that not inconsistent?" Surely the hunter's query concerning
the consistency of his critic's position is here quite a justified move in argu­
ment.
However, in the original argument we addressed ourselves to (the
sportsman's original rejoinder), the propositions which made up the allega­
tion of inconsistency should be carefully examined. What the sportsman
was claiming in that case, is that there was a circumstantial inconsistency
between the following two propositions.
1. The critic criticizes the killing of game for amusement.
2. The critic himself eats meat.
Is this really an inconsistency? One author, DeMorgan (1847, p.265),
pointed out that there is merely the superficial appearance of inconsistency.
A celebrated writer on logic (Whately) asserts that no one who eats meat
ought to object to the occupation of a sportsman on the ground of cruelty.
The parallel will not exist until, for the person who eats meat, we substitute
one who turns butcher for amusement, according to DeMorgan.
What DeMorgan is saying is that there is a certain lack of proper paral­
lel here. It might be inconsistent to oppose hunting if one is in fact a hunter,
but there is nothing directly parallel in the case of the critic who eats meat
yet decries the hunter for killing game for amusement. In short, decrying
hunting while eating a juicy steak is not directly circumstantially inconsis­
tent.
DeMorgan made good points in noting that the critic's argument does
not express a direct circumstantial inconsistency against the hunter's posi­
tion and that the hunter would be quite justified in responding by pointing
this out. Yet on the other hand, in all dialogical fairness, it must be recog­
nized that there is in fact some connection between the activities of hunting
and eating meat. After all, the practice of eating the flesh of animals surely
is a main economic factor in making the slaughtering of animals a profitable
activity. Therefore meat-eating is certainly a way of contributing to the
extent of the practice of the killing of animals by humans. Surely therefore,
meat-eating is not entirely unconnected to hunting.
224 INFORMAL FALLACIES

Yet we must observe also that the connection is not by any means a
tight one. There are indeed other ways of obtaining meat for human con­
sumption than hunting. Moreover, hunting could still be perceived by its
exponents as a pleasurable or desirable activity even if the meat thereby
obtained was not used for human consumption. So while there is a connec­
tion between the two propositions at issue, the precise nature of that con­
nection is disputable and is in fact, also at issue.
Ultimately, the resolution of an ad hominem dispute then must turn
around the precise nature of this connection in a particular case. One can
see therefore why the use of the sportsman's rejoinder as an example of the
circumstantial ad hominem fallacy in the traditional textbooks has tended
towards a certain superficiality. Whether or not the sportsman does argue
fallaciously requires a good deal of further analysis of the specifics of the
case by further dialectical questioning of the participants.
However, despite the lack of specifics, given the details of the dispute
the way the example stands, can we say that either participant has commit­
ted a fallacy? Certainly DeMorgan was right to point out that the
sportsman's initial argument was fallacious to the extent that the inconsis­
tency he alleged did not really obtain. In effect therefore, the sportsman
simply got his propositions wrong. If he meant to say that the critic was
directly inconsistent, he was certainly mistaken in this claim as presented,
and the recognition of the slipperiness of this particular move in argument
could certainly be labelled an important type of ad hominem fallacy. The
sportsman's allegation of direct inconsistency could not be justified by the
given evidence, and thus the sportsman's attempt to mount a conclusive ad
hominem refutation was unsuccessful. Thus he committed a form of ad
hominem fallacy by doing such a clumsy job of mounting his ad hominem
allegation.
In chapter two, section two, we encountered a kind of reasoning called
practical inference. This type of inference takes the following forms, where
X is an agent and A and  are things carried out by the agent (events or
outcomes).
X intends to do A.
To do A, it is necessary to do B, as X sees it.
Therefore, X sets himself to do .
X intends to do A.
To do A, it is sufficient to do B, as X sees it.
Therefore, X sets himself to do .
 ARGUMENTS AGAINST THE PERSON 225

Both these forms of inference have to do with an agent's intentions or plans

in carrying out a sequence of actions.
But there are several other conditions that have to be met for the prac­
tical inference to be more complete. If there is more than one way available
to carry out A then X need not necessarily choose  as part of a reasonable
plan of action. However, the premiss may be added that  is the only way
available in the circumstances, or that  is the best way. Additionally, side
effects should be taken into account. If carrying out  could have some very
bad side effects, an agent might rationally question whether these outweigh
the positive value of carrying out A for him. So in some cases, a premiss
weighing the value of A against the possible consequences of  should be
considered by a rational agent.
Having added these additional premisses, there comes a point when
practical reasoning in a particular situation can be complete in the following
sense. If an agent accepts the premisses but refuses to act so as to carry out
the event described the conclusion, he is open to a challenge of inconsis­
tency in his intentions and actions. For example, suppose that a person
intends to eat some food and he considers that a large steak on the table in
front of him is a bit of food. Moreover, it is clear to him that this steak in
front of him is the most easily available bit of food, and it is the food he
thinks best at the moment to satisfy his appetite. Moreover, he has no
reason to believe that there will be any harmful or bad side effects of eating
the steak, and that it would be in no way illegal, immoral, unhealthy, or
whatever, for him to eat the steak. But suppose he does not begin to eat the
steak, and instead just sits there filing his fingernails. Is he being inconsis­
tent?
According to the theory advocated in Walton (1985), we could chal­
lenge this man's practical reasoning by charging him with being practically
inconsistent. We mean by this criticism that there is an ostensible incon­
gruity between the man's intentions and his actions, relative to his percep­
tion of the situation he is in. Such a pragmatic inconsistency shifts the bur­
den of proof onto the person criticized to justify his position. Was the situa­
tion not really as we thought he perceived it? Or were his intentions differ­
ent than we thought? Barring a good response to the challenge, the man's
actions may plausibly be taken to represent an inconsistency of position.
The vulnerability of a practically inconsistent argument to ad hominem
criticism is well known to lawyers who are skilled in cross-examination of
witnesses or defendants as to their actions. When a witness admits to acting
226 INFORMAL FALLACIES

in a manner inconsistent with his own intentions and perceptions of a situa­

tion, exploring and pinning down the inconsistency is a known way to
impugn the plausibility of the testimony of that witness.
In the following case, a police officer was a complaining witness against
a defendant accused of driving through a red light and not giving the police
car the right of way. The police officer's intentions were made clear by his
admission that he was attempting to rush to the aid of a fellow officer who
was in danger. However, he was impeded by the defendant, who drove into
his police car.
A sequence of questions used to cast doubt on the police officer's tes­
timony is outlined by Schwartz (1973, p.2014f.) as a sample dialogue. The
attorney opens with a series of questions to establish that the police officer
was responding to an emergency and was hastening along a clear, dry
street, trying to make every moment count. The attorney's dialogue with
the police officer then took the following form (p.2105).
Q. So that you had a clear path open ahead of you as you
approached the intersection?
A. Yes.
Q. Your siren was blowing full blast?
A. It was.
Q. Your lights on top were rotating?
A. They were.
Q. You were intent on getting there just as soon as you possibly
could?
A. Yes.
Q. Yet, I noted that you testified that your speed was only five miles
an hour — was that your testimony?
A. That was at the intersection only.
Q. What had been your speed before you came to the intersection?
A. About 40 miles per hour.
Q. Did you slow down to 5 miles an hour because you suddenly
changed your mind about getting there in a hurry?
A. No.
Q. Did you insist that while rushing to the aid of a fellow officer,
with the path clear in front of you, your siren on full blast, the
lights rotating, and with the traffic light in your favor, you reduced
your speed at that intersection from 40 to 5 miles an hour?
 ARGUMENTS AGAINST THE PERSON 227

A. Yes, I did.
Defense Counsel: That will be all.
The effect of this line of questioning on the deliberations of a jury would be
quite pointed. By revealing such a presumptive inconsistency in the position
upon which the police officer apparently acted, it would throw grave doubts
on the plausibility of his testimony. If his intentions and the circumstances
of his actions were as he conceded, then there could be no obvious reason
why he should have been driving slowly instead of driving quickly. Short of
our being given some such reason or additional fact about the situation, we
can scarcely draw any other conclusion than the practical reasoning, taken
as a whole sequence, is inconsistent. This, of course, leads us to strongly
suspect that it was the police officer, and not the defendant, who went
through the red light.
Thus practical reasoning can play an important role in shifting the bur­
den of proof in dialogue. The skilled use of dialogue can bring an ad
hominem criticism to bear against the position of an agent.
One main problem with nailing down ad hominem criticisms — and
with defending your position against them, for that matter — is that practi­
cal reasoning often comprises a complex sequence of actions. And any
sequence of actions can be described in many different ways. Hence there
are many openings to check in evaluating any ad hominem attack on the
basis of circumstantial inconsistency.

3. Evaluating Ad Hominem Disputations

The complexity implicit in the dialectic of the sportsman's rejoinder
suggests that in any allegation of circumstantial ad hominem there are at
least ten check points which should be carefully clarified in evaluating the
dispute. First, what are the propositions that are alleged to be inconsistent?
Rather than concentrating on the personalities of the arguers it is better to
concentrate on the propositions that make up the alleged inconsistency and
first of all clearly identify them. Second, who advanced the allegation of
inconsistency? We presume here that there is an identifiable game of
dialogue in progress and that one participant alleges that another has com­
mitted the inconsistency.
Third, we must ask — who was alleged to have committed the inconsis­
tency? We have to be careful here that the perpetrator of the inconsistency
could be an individual, a group or part of some group of individuals. We
228 INFORMAL FALLACIES

also have to be careful that this class is clearly specified throughout the
argument, because sometimes shifts can be made. For example, supposing
"the press" is collectively accused of making untruthful statements and then
it is alleged that "the press" has a certain moral obligation to the truth
because of its position in society. It could be argued then that there may be
a certain circumstantial inconsistency on the part of the press as a collective
group but we have to be careful to ask the following question here. Is this
some particular group of members of the press that are being alleged to be
inconsistent? That is, who in particular, which individual or which groups,
were alleged to have made the false statements at issue? Often in longer
extended arguments, ad hominem allegations fail to clearly and consistently
specify the specific group, and therefore the ad hominem allegation must be
judged to fail for this reason.
Fourth, we have to ask what type of inconsistency may be involved. As
we saw in the sportsman's rejoinder, it may not be truly logical inconsis­
tency but rather an action-theoretic inconsistency. Sometimes also the alle­
gation is one of an indirect inconsistency. The critic argues that there is
some connection between some action of the opponent and some proposi­
tion that the opponent has put forward. However, the connection here may
not be so much logical in nature as more causal or action-theoretic, and
therefore we must be careful not to confuse different types of allegations of
inconsistency.
Fifth, once some inconsistency has been specified, can it be really
proven that the propositions are inconsistent? Sixth, if the inconsistency
amongst the set of propositions can be proven, has a substitution been
made as in the DeMorgan warning about the sportsman's rejoinder? We
have to ask whether there is merely a near parallel as opposed to direct
inconsistency between two propositions.
Seventh, if there is no direct logical inconsistency between the proposi­
tions identified, is there rather some indirect causal linkage between the
propositions that make up the set alleged to be inconsistent?
Eighth, if there is no inconsistency at all, then should we say that the
accuser himself is guilty of committing an ad hominem argument by a
clumsy mounting of the ad hominem refutation? Ninth, if there is not a log­
ical inconsistency, but only a pragmatic one or causal connection between
the propositions, then how serious a flaw is this in the arguer's position?
Perhaps only further dialectical moves can resolve this question.
Tenth, does the accused have a chance to respond to the allegation of
 ARGUMENTS AGAINST THE PERSON 229

inconsistency? We saw earlier that the ad hominem attack often has the
effect of stopping argument altogether, and that therefore it is always quite
important, because of the serious nature of any ad hominem accusation,
that the accused party have some chance to respond to the allegation and
perhaps retrieve the argument.
So to return to our analysis of the sportsman's rejoinder, we should
now ask whether there was indeed a pragmatic inconsistency to be found in
the hunter's ad hominem allegation. We saw previously that there is some
connection between hunting and meat-eating. Therefore, perhaps the
sportsman could be on strong ground if he were to pursue this interconnec­
tion. Perhaps then some legitimate form of argument could be dialectically
extracted from what the sportsman says. Of course, as we saw above, there
is no tight logical connection between hunting animals and eating meat.
Therefore much depends on our reconstruction of the sportsman's argu­
ment on what sort of connection he might have had in mind. Did he mean
to argue that there was a tight logical connection and thereby conclude that
the critic is inconsistent? If so, his argument as first presented would defi­
nitely constitute a form of ad hominem fallacy.
However, perhaps that is not what the sportsman meant to say. Sup­
pose that in subsequent dialectical interchanges — if we could continue the
argument further between the hunter and his critic — the hunter manages
to get the critic to agree to the following set of commitments.
1 . T h e action of eating the flesh of animals is a sufficient condition of the
proposition that animals are killed.
2. Something I (the critic) did brought it about that animal flesh was
eaten.
3. I (the critic) did not bring about the killing of any animals.
Now this set of propositions is demonstrably inconsistent. It is not necessarily
logically inconsistent, but it could be shown to be pragmatically inconsistent
in the sense that the description of the actions collectively constituted by
the set of three propositions demonstrably leads to an impossibility. The
first proposition states that one action is a sufficient condition of another
action. The second proposition asserts that the first action is attributable to
the critic. Therefore, by a form of modus ponens we can conclude that it
follows that an action of the critic brought about the death of some animals.
However, by asserting the third proposition, the critic denies that he
brought about the death of some animals. This last proposition is the nega-
230 INFORMAL FALLACIES

tion of the preceding one. Thus we can see how the set of three proposi­
tions above does lead by steps of logical reasoning to an inconsistency.
The principle of reasoning by which we derived this contradiction,
namely modus ponens, is a rule of classical logic. But the type of inconsis­
tency involved is not a directly logical inconsistency. Rather it is one that
has to do with the bringing about of certain actions and the question of how
some actions are related to other actions. The inconsistency is what we
might call a pragmatic inconsistency — one that has to do with relationships
amongst collections of actions attributed to some individual. If 1., 2., and 3.
above really constitute what the critic is saying, then the hunter is correct to
point out that the critic, while he may not be directly logically inconsistent,
is nevertheless inconsistent in an action-theoretic way. His set of proposi­
tions that he asserts can be reduced, through a description of the actions
they constitute, to a direct logical contradiction.
However, from our original description of the argument between the
sportsman and the critic can we really fairly say that in fact the critic is
asserting precisely these three propositions? The answer is 'no,' we cannot.
Only further dialectical interchanges between the two disputants could
clarify whether in fact this is really the argument the critic means to put for­
ward. If subsequent dialectical interchanges do detemine that this is in fact
the critic's argument however, the sportsman is then in a position to cor­
rectly turn the tables on the critic. It is not the sportsman who has commit­
ted the ad hominem fallacy (although he may have committed an ad
hominem of a different sort, as we saw above), but it is the critic who has
committed himself to a certain action-theoretic inconsistency, and his move
in so doing may also be called a form of ad hominem argument.
The long and the short of this dispute is that any ad hominem allega­
tion must be carefully looked at so that the specific dialectical context can
be exactly determined and the precise propositions at issue attributable to a
certain disputant can be determined. Only then can we analyze whether in
fact there has been a fair allegation of inconsistency made or not.

4. Four Types of Circumstantial Ad Hominem

We have found that there can be different forms of the ad hominem
argument, that in a dispute one party can accuse another of ad hominem
and yet depending on how you reconstruct the argument from the dialecti­
cal point of view, it could be that either or both parties is guilty of the same
 ARGUMENTS AGAINST THE PERSON 231

or different forms of the ad hominem fallacy. Let us therefore try to sum up

how in a dialectical situation like this there can be various different forms of
ad hominem fallacy committed by one party or another in the dispute.
There are four different types of possible illegitimate moves of
argumentation that could be construed as the circumstantial ad hominem
fallacy. We are proposing that an ad hominem type of argument is not
necessarily in itself fallacious, and that it can be quite a legitimate move in
argument to find an inconsistency, whether logical or pragmatic, in an
opponent's argument. However, in making this type of move, the attemp­
ted refutation may be carried out badly, and the attacker may then sub­
sequently commit one or the other of four different types of fallacy.
In the first type of fallacy the attacker finds merely the appearance of
inconsistency but does not prove there is really an inconsistency either of
the pragmatic or the logical sort. This is DeMorgan's point that an attacker
may find two propositions that are parallel to an inconsistency, or which
may be related to each other in such a way that an inconsistency could pos­
sibly be derived from them by further dialogue, but unless the attacker
actually determines that there is a logical inconsistency by the rules of the
game of dialectic and by particular propositions that are really assented to
by the opponent in argument, his rejection of that argument commits an ad
hominem fallacy. The refutation is over-hasty and unsuccessful. It is all too
easy to suggest that there may be an inconsistency, and that in itself is
sometimes enough to persuade an audience to reject somebody's position in
argument. Nevertheless we must look very closely to see whether, given the
rules of the game, there really is a determinable inconsistency according to
those rules. So if the attacker or the audience rejects somebody's argument
merely because there is the superficial appearance of inconsistency, but not
the reality of it, this is the first form of the ad hominem fallacy.
In the second form of fallacy, the attacker finds an inconsistency, and
then on the basis of this finding, rejects the defender's conclusion per se.
This form of the fallacy raises the general question of how to handle con­
tradictions in games of dialogue. In some games of dialogue, the object of
the game is to trap the opponent in a contradiction. Therefore if one disput­
ant succeeds in doing this, the game is over and that disputant wins by the
convention of the game. However, in other games the win-strategy of the
game may not be defined as that of trapping the opponent in inconsistency,
and there may be means for handling inconsistency in such games. For
example, in some games of dialectic, if a disputant finds himself in the posi-
232 INFORMAL FALLACIES

tion of having a demonstrable inconsistency in his commitment-store he

may be able to retract one or the other of the commitments in order to
eliminate the inconsistency. For example, several of the games designed by
Hamblin do permit the retraction of commitments for this very purpose. In
such a game then, if a disputant is found to have an inconsistency in his
commitment-set, it does not mean that he has thereby automatically lost
the game, nor does it mean that his thesis is thereby refuted.
However, we have already previously noted the rhetorical power of a
finding of inconsistency in someone's internal position by an opponent in
argumentation. There is nothing more ridiculous than being caught in
inconsistency and audiences find this most amusing. Sometimes it can even
be suggested that inconsistency is a sign of mental instability or insanity. At
any rate inconsistency of commitments is often taken to be some sign of a
pronounced lack of logical acumen, and as we have seen, the pointing out
of it has such a damaging impact as a refutation that it may have the effect
of stopping the argument altogether or shifting it off topic. However, it may
be quite fallacious to conclude that the opponent's thesis is itself false, sim­
ply because some pair of propositions that he has accepted, that may even
not be directly related to his conclusion, turn out to be not consistent with
each other. In fact if the pair of propositions that are inconsistent are not
directly related in any way to the conclusion, the finding of inconsistency
may not be seriously damaging to this participant's over-all argument at all.
Nevertheless, it is very easy to commit the ad hominem fallacy by automat­
ically rejecting this participant's whole position as being illogical, and con­
sequently rejecting the thesis that he is supposed to prove without further
consideration. This, then is the second form of ad hominem fallacy that may
occur.
The third form of the fallacy occurs where the attacker finds a pragma­
tic inconsistency, but then rejects the defender's argument as logically
inconsistent without proving that it is logically inconsistent. We have seen
that there is a difference between action-theoretical and pragmatic inconsis­
tency on the one hand, and logical inconsistency on the other, the latter
being a stronger type of inconsistency.1 The former is only reducible to the
latter by additional steps of argument. Therefore, one possible fallacious
move in argument could be to declare the defender's position as logically
inconsistent when really the inconsistency is a matter of what he is commit­
ted to by his actions as opposed to the commitments he makes by the prop­
ositions that he asserts.
 ARGUMENTS AGAINST THE PERSON 233

Often a pragmatic inconsistency can be defended. For example, Ber­

trand Russell was once accused of being sympathetic to the Soviet Union in
his comments on international affairs, and then later at some subsequent
point, taking a very hard line on international policy towards the Soviet
Union. However, Russell defended his actions on the grounds that the
obtaining of nuclear weapons by the Soviets had changed the situation and
subsequently made a change in policy towards the Soviets a reasonable
move.
In a fourth form of ad hominem fallacy, the attacker finds some con­
nection between two or more descriptions of actions attributed to the
defender, but fails to show the requisite connections between these descrip­
tions of actions in order to demonstrate the existence of a pragmatic incon­
sistency within them. Therefore, in this fourth form of fallacy, the attacker
concludes that the defender is inconsistent without carefully filling in the
steps needed to adequately establish the inconsistency.
To sum up, we can see that there are at least four different types of ad
hominen fallacy2 and that there are various complications of ad hominen
attack and defence, depending on the particulars of a given dialectical situ­
ation.

5. Rhetorical Context of Ad Hominem Attacks

The fact that the ad hominem argument is a dialectical phenomenon
means that one always has to be very careful in looking at the dialectical
context of the argument to try to see who is making the ad hominem allega­
tion, who is being accused of the ad hominem allegation, and what are the
propositions that make up the basis of this allegation. Finally, one has to
carefully look at these propositions to see what sort of logical relationships
obtain amongst them. One needs to find out whether these are alleged to be
logical relationships or causal or action-theoretic relationships. And then
one has to try to ascertain, perhaps by asking further questions to one or
both participants, whether enough steps have been filled in by the attacker
to justify a sound basis of ad hominem refutation. If not, the attacker may
have committed an ad hominem fallacy. However, if the attacker has done
a good enough job of filling in the steps of his argument in order to show
that there is a genuine inconsistency, then the defender may be guilty of an
ad hominem fallacy. So the ad hominem fallacy is a difficult one to adjudi­
cate when one looks at some of the examples given in the textbooks. Com­
monly there is not enough information given about the dialectical context
234 INFORMAL FALLACIES

of the dispute in order to provide adequate evidence to determine which

party has truly committed an ad hominem fallacy.
Suppose, for example, that a philosophy professor makes strong
attacks on physicians' rights to make paternalistic decisions on behalf of
their patients. Suppose further that a physician replies that it is entirely nat­
ural for a philosopher to feel some degree of resentment and frustration
because he is in the relatively powerless position of being a pure theoreti­
cian. Suppose then the doctor queries whether there might not be an
unconscious wish to strip powerful physicians of their authority. Could this
philosopher's criticism really be motivated by his own power-seeking fan­
tasy and his jealousy of the great powers of physicians? The move here in
argument by the physician qualifies as an ad hominem attack because the
personal circumstances of the philosophy professor are brought in as rele­
vant to his claim about paternalistic decision-making on the part of physi­
cians. However, could not the doctor defend his argument as follows.
My argument never positively affirms that what the philosophy professor
says is wrong. I am merely speculating on the psychological basis of the
philosopher's reasoning.3
The problem with this response, however, is that there may be a good deal
of harm done more by way of innuendo than by precisely articulated propo­
sitions. An innuendo is not necessarily an argument, a precise set of propo­
sitions that one can pin down to attribute specifically to a participant in
argument, but it may nevertheless have quite a substantial effect on a par­
ticular audience in deciding how they will ultimately act on the basis of the
argument. Here we are at the pragmatic edges of the ad hominem fallacy
and we have some difficulty in nailing it down.
The doctor's argument suggests very strongly that his opponent's argu­
ment could be prompted by the resentment of a powerless theoretician. In
effect therefore, his opponents are being characterized as unqualified and
frustrated meddlers, interfering with doctors because of their resentment of
the doctors' powers. The suggestion conveyed by this argument therefore,
is an extremely powerful and captivating one from a rhetorical point of
view. Many readers would, with justification, be inclined to accept the
proposition that arguments concerned with public affairs are very likely to
be prompted by the arguer's own professional interests and affiliations.
And in fact professional groups notoriously do tend to lobby for their own
special interests on such topics. Thus the problem posed by such arguments
is that in the marketplace of conversational disputation, it may be quite dif-
 ARGUMENTS AGAINST THE PERSON 235

ficult to precisely pin down the ad hominem allegation4 and therefore

adjudicate fairly upon it.
The problem is that the dialectical circumstances are incomplete. The
nature of the argument and the propositions that make it up are unclear.
The conclusion that is supposed to be inferred from such quasi-arguments is
never made explicit yet they can be phrased in such a way that undoubtedly
they will have their intended affect, and the target audience of the argu­
ment will be strongly influenced by what is implicitly contained in the pre­
misses. Therefore, we are back to the level of the debate, where it is dif­
ficult to regulate or adjudicate upon arguments because of their essentially
subjective nature. Here then we arrive at the borderline between dialectic
and the rhetorical debate.
In the rhetorical debate, the outcome is decided not so much by the
conformity of the participant's requirements, but by the vote or psycholog­
ical outcome of some target audience. Hence, in the debate, the decision
has to be dependent upon purely psychological and rhetorical factors, on
the impact of the participant's arguments upon the particular subjective
impressions of that audience. Clearly, many audiences would be strongly
moved by the physician's ad hominem argument, especially if there is some
antecedent proclivity on their part to be inclined in the direction of sym­
pathy with the physician's point of view. Here then is the problem of trying
to study or analyze the ad hominem fallacy at the level of the debate or con­
versational quarrel. What is or is not fallacious may not be precisely analyz-
able without going back to the participants themselves to reconstruct more
clearly what is the precise argument and what are the precise rules of the
game. In many cases it may not even be clear or determinable whether a
participant has in fact offered an argument in the sense of a set of proposi­
tions. In such a case, even at the outset we are scarcely in a position to
adjudicate upon whether this participant has committed a fallacy.
The lesson then is that the analysis of what is fallacious about an ad
hominem argument can only take place at the dialectical level, where
moves and outcomes are determined by objective rules, and where relation­
ships can be precisely determined to obtain amongst a set of propositions by
questioning the participants. At this level, the winner or loser of the argu­
ment need not be fully determined by the psychologically varying reactions
of some particular audience.
At the dialectical level then, we can approach the analysis of a
sequence of argumentation and ask a number of questions. First, is it an
236 INFORMAL FALLACIES

argument, or merely a question or some other form of non-statemental

move? Second, what are the propositions in the argument? What sort of
allegation is made in regard to them? Does the opponent have a chance to
reply? Does the sequence of interchange between the participants qualify as
being a win-strategy for one or the other participant? Can we query the
opponents further to see precisely what is the sequence of argumentation
that they have in mind? Can a graph of the argument be constructed so that
we can get a precise account of its overall flow and direction and fill in the
missing premisses and links of argument? What are the rules of argument,
and have these been precisely determined in advance by the participants?
Once a number of these dialectical factors have been filled in and clarified,
then much that is both objectionable and elusive from a point of view of
criticisms — like the one attributed to the physician in this example — can
be clarified or at least forced out into the open where it can be processed
and eventually confronted and adjudicated upon in a way that is fair to both
parties.

6. Positional Defensibility
In many ad hominem allegations the cited inconsistency is of a deontic
(normative) character. The accusation amounts essentially to the claim that
some participant acts in such a manner to bring about some state of affairs,
but also claims that state of affairs ought not to be brought about. Thus the
inconsistency involved is not a directly logical inconsistency, but is what we
might call a deontic pragmatic inconsistency,5 that is, an inconsistency
which has to do with norms and actions. In this kind of criticism the
attacker is essentially saying this: "You [the person I am criticizing] say that
such-and-such ought to be brought about; but you yourself, in your own
practice, don't act in such a way as to bring this about at all. Therefore,
your position is inconsistent, and you are a hypocrite." Now what we
should ask here is whether such a deontic pragmatic inconsistency in one's
position should really be a damaging deficiency to one's argument in a
game of dialectic.
First of all, we should note that recommendations from a position that
is deontic-pragmatically inconsistent are extremely vexing to persons who
are themselves forced to act in a manner consistent with the pronounce­
ments of one who fails himself to act in that very manner. We are outraged
for example, when politicians tell us we have to do certain things like
"tighten our belts" and live in accord with reduced economic resources, and
yet themselves fail to do this by their own high spending, or by unrealistic
 ARGUMENTS AGAINST THE PERSON 237

increases in their own salaries.6 Such inconsistency can indeed be immoral,

and can certainly appear to directly violate fundamental ethical principles
of fairness and democratic government. Moreover, such inconsistency in
one's actions can be evidence of callousness, moral indifference and stupid­
ity. However, we must ask whether the fault there is a purely ethical error
or whether it can also be evidence of the existence of some logical error that
should lead us not to accept positions in games of dialectic that contain such
an inconsistency.
The first problem is to define what counts as an opponent's position in
argument. For example, because one is female it hardly follows that one
adopts a "feminist position." What counts as the so-called "Marxist posi­
tion" may be quite open to disputation even among the Marxists them­
selves. Therefore we must be extremely careful in presuming that we may
confer some "position" on an opponent in argument. We should recall the
account of position in argument in the games ABV and CBV.
Let us repeat that a position should be defined in a game of dialogue as
a set of propositions that are dark or light-side commitments of a partici­
pant in the game. Accordingly, in order to precisely define position, one
needs to specify the game that one has in mind.
A wider way of defining position could be as follows. A position is a set
of propositions that form a basis for acting when one is formulating a plan
of action. From this point of view a position could be thought of as a sort of
code or body of propositions that represent outcomes one plans to bring
about and deliberates to achieve by means of one's actions.
So construed, a position may equally well be the code of a group of
agents. For example it may be a statement made by a group of individuals
who want to express their resolve to support nuclear disarmament, or it
may be a statement of commitment to a professional ethic like the Hyppoc-
ratic Oath. One the other hand, a position may be an individual's statement
of his own goals, an expression of where he or she stands on an issue.
Given the concept of a position, we can then proceed to elucidate
further what is fallacious about the circumstantial ad hominem inconsis­
tency of the pragmatic sort. If a proposition is part of some participant's
commitment, we may sometimes infer that this proposition forms a basis
for an action-plan that this participant has formulated. Let's say then that a
participant in dialogue has committed himself to bringing about a particular
proposition as part of a plan he has advanced. But if it can be shown that
this participant as a matter of fact brings about the negation of this very propo­
sition, there is a sense in which the participant's internal position is inconsis-
238 INFORMAL FALLACIES

tent. If so, the participant's plan of action is incoherent in just precisely this
sense. His goal of deliberation is to bring about the truth of a certain prop­
osition p, but in his own conduct he deliberately acts so as to bring about
the negation of this very proposition p. Therefore, his plan of action, his
collective position as a whole, is inconsistent. The question then can be
reformulated as follows.
What is wrong with an inconsistent plan of action? The answer is that
an inconsistent plan of action can never be carried out. A set of proposi­
tions that are logically inconsistent are globally destructive in the sense that
they collectively imply any proposition you like. However, a positional
inconsistency results in a non-directive plan. It allows you to follow any
course of action that happens to be related to your position. It thereby
becomes non-directive by failing to inform you how to make a rational
choice between two outcomes. In short, a positional inconsistency makes
your position useless as a basis for action. Thus, the participant's position
that turns out to be pragmatically inconsistent exposes a deep irrationality
in that position, and hence the argument that is based on that position
should be questioned, and also perhaps rejected if the inconsistency cannot
be remedied.
It follows then that positional inconsistency can be determined as a
legitimate basis for the criticism of an argument and that therefore ad
hominem criticisms that one's internal position is inconsistent are some­
times a correct form of argument.
However, as we have seen, there are many ways in the course of a
dialectical disputation that such an accusation can go wrong, and when one
of these occurs, we have the ad hominem fallacy, in one form or another.
However, even a successful refutation of the ad hominem sort in argumen­
tation is rarely final. Rather a successful refutation may be defined in rela­
tion to its objective of shifting the burden of proof onto the participant who
is attacked to remove the inconsistency, and in practice many such allega­
tions need a lot of tightening up before it can be conclusively determined
which party is in the wrong or right.

7. Conclusion
We conclude that much of the prevailing Standard Treatment of ad
hominem arguments rests upon a superficial assumption that ad hominem
allegations are always per se fallacious. By now it is clear that we cannot
accept this approach. Instead we have to very carefully distinguish between
 ARGUMENTS AGAINST THE PERSON 239

an ad hominem attack and an ad hominem refutation. In an ad hominem

attack a disputant brings forward a set of propositions alleged to belong to
the position of his opponent. These propositions are very often things said
to be brought about within the context of an action plan of the attacked
participant. An ad hominem attack can then become a successul refutation
if the set of propositions alluded to can be shown by the attacker to be
inconsistent according to the rules of the game of dialogue the participants
are engaged in.
Having distinguished between an ad hominem attack and an ad
hominem refutation, we should now review what constitutes the ad
hominem fallacy. To review, we should recall that a refutation can go badly
wrong in a number of characteristic ways we have catalogued. For example,
the attacker may determine that two propositions are part of his opponent's
position, but then instead of clinching a direct logical inconsistency between
these two propositions, he may only find some indirect linkages between
the two that need to be filled in by further steps. Or in another way of com­
mitting the ad hominem fallacy, he may simply mis-describe the proposi­
tions in such a way that they are parallel to an inconsistency but not, so
described, directly inconsistent. In yet a third form of ad hominem fallacy,
the attacker may secure a legitimate contradiction in the opponent's posi­
tion, but then by non sequitur reasoning declare that one member of the
inconsistent set must be in itself false.
We must be very clear then to keep in mind that an ad hominem attack
is not always a fallacy. Indeed, we have claimed that under the right condi­
tions an ad hominem attack can turn out to be a successful ad hominem
refutation and therefore essentially a correct form of argument. Yet, if cer­
tain conditions fail to be met, the refutation itself can be sophistical and
turn out to be an ad hominem fallacy. In treating many actual disputations
of natural conversation, it is often best to ask not whether the argument is
fallacious or not as it stands, but whether it is vulnerable to ad hominem
criticism. That is, in the case of an attack we should sometimes ask not
whether the argument is fallacious or incorrect as it stands, but whether it
can be filled in in such a way that it could eventually turn out to be tenable
as an allegation and hence constitute a successful ad hominem refutation.
Such judgements must therefore often be made in the context of the debate
and quarrel as conditional upon the further continuation of the dialectical
sequence.
240 INFORMAL FALLACIES

NOTES

1. This is not the place to attempt to systematically define action-theoretic inconsistency, but
the reader may like to consult some of the literature on the topic. See G.H. von Wright, An
Essay in Deontic Logic and the General Theory of Action, Amsterdam, North-Holland, 1968.
2. These four forms of ad hominem fallacy are identified and more fully analyzed in Walton
(1983).
3. An actual example of a disputation from which this account is drawn is fully presented and
documented in Walton (1983).
4. Hamblin (1970) refers to the general problem in practice of "nailing down a fallacy."
5. This term originates in John Woods and Douglas Walton, 'Ad Hominem,''The Philosophi­
cal Forum, 8, 1977, 1-20.
6. An example of this sort is presented by Trudy Govier, 'Worries about Tu Quoque as a Fal­
lacy,' Informal Logic Newsletter, 3, no.3, 1981, 2-4, and is further discussed in Walton (1985).
 CHAPTER 10: EQUIVOCATION

Among Aristotle's fallacies within language, the fallacy of equivoca­

tion is that fallacy that has to do with an ambiguous term in an argument.
Where it is a construction of terms in a sentence, and not just one term in
the sentence, that may be taken more than one way, then the fallacy is said
to be that of amphiboly. As we have seen, it has never been established that
amphiboly is truly a serious fallacy, in the sense of being a significant source
of mischief or confusion in realistic argumentation. But one example, if it is
a case of amphiboly, seems to be a serious and reasonably realistic sort of
error. The argument, 'Everything has a cause, therefore there is something
that causes everything' could be given the following form in classical first-
order logic: (Vx) () ( causes χ), therefore (3y) (Vx) (y causes x). But
we will say no more about amphiboly. For most of what we will have to say
about equivocation in the sequel applies equally well to amphiboly, in gen­
eral outline. And there is very little doubt that equivocation is a serious fal­
lacy, well worth detailed investigation.
Equivocation is said to occur where an ambiguous term occurs in an
argument that appears to be valid. What is meant by 'appears to be valid,'
in the context of equivocation, is that the sentential structure of the "argu­
ment" resembles that of a valid form of argument. For example, the follow­
ing argument appears to be valid, in this sense.
(1) Everything that runs has feet.
Some rivers run.
Therefore, some rivers have feet.
Certainly the sentences of this "argument" have a syntactic structure that is
analogous to a valid form of argument in classical logic: (Vx) (x runs ⊃ 
has feet), (Hx) (x is a river  χ runs), therefore (3x) (x is a river   has
feet). But there is more to the fallacy of equivocation. For we can have syn­
tactic structures analogous to valid arguments where the ambiguous term
'runs' also occurs. But these "arguments" would not be ones we would
probably want to call fallacies of equivocation.
242 INFORMAL FALLACIES

(2) Animals run.

Rivers run.
Everything that runs moves.
Therefore, animals and rivers move.
(2a) Everything that runs goes north.
Some rivers run.
Therefore, some rivers go north.
But perhaps 'run' is only ambiguous in (1) and not in (2) or (2a). Or if a
term is ambiguous, is it ambiguous wherever it occurs? If so, then there are
some arguments that contain ambiguous terms and have a syntactic struc­
ture that is a valid form, but do not commit the fallacy of equivocation.
And if so, then the characterization of fallacies of ambiguity as arguments
that appear syntactically valid in form but contain words that, either singly
or in combination, can be understood in more than one sense, is somewhat
too broad to pin down exactly what is fallacious about equivocation.
What is fallacious as an equivocation about (1) is not just the surface
validity and ambiguity in the argument, but a contextual shift. To make the
first premiss plausibly true, we are tugged to interpret 'runs' as meaning
'ambulates.' To make the second premiss plausibly true, we are tugged to
interpret 'runs' as meaning 'moves along' in a sense that counts flowing as
running. That disambiguation clashes with the one needed to make the first
premiss true. So equivocation involves contextual shift as well as validity
and ambiguity.

1. What is Equivocation?
What then is the fallacy of equivocation? Let's begin with one elemen­
tary, yet basically important lesson. It has often been cogently pointed out,
for example by Hamblin (1970) that there is a difference between ambiguity
and equivocation. Ambiguity need not be fallacious. Indeed, on the
assumption that a fallacy is a fallacious argument, ambiguity cannot be fal­
lacious. For ambiguity is a property of sentences rather than arguments. At
any rate, the crux of this basic point is that the fallacy of equivocation
resides in the fallacious deployment of ambiguity in arguments. But what
exactly do we mean by 'deployment,' and how is such fallacious deploy­
ment affected in the use of arguments?
A clue is given in Woods and Walton (1979), where a characterization
of equivocation is based on Quine's insight that a contextual shift between
 EQUIVOCATION 243

sets of sentences or propositions is involved. The traditional example of

equivocation is this familiar old favorite of the texts.
(3) The end of a thing is its perfection.
Death is the end of life.
Therefore, death is the perfection of life.
According to the traditional analysis, 'end' can mean 'goal' or 'termina­
tion.' Hence the one argument is really four.
(4) The goal of a thing is its perfection.
Death is the goal of life.
Therefore, death is the perfection of life.
(5) The termination of a thing is its perfection.
Death is the termination of life.
Therefore, death is the perfection of life.
(6) The goal of a thing is its perfection.
Death is the termination of life.
Therefore, death is the perfection of life.
(7) The termination of a thing is its perfection.
Death is the goal of life.
Therefore, death is the perfection of life.
Of these four arguments, (4) and (5) are deductively valid, but (6) and (7)
are invalid. However, only (6) has true premisses. (4) and (5) each have
one premiss that is untrue (or at least implausible). In (7) both premisses
are false (or, at any rate, implausible). Hence the person to whom the con­
stellation of argument (3) is directed is given a potentially mischievous
choice. He can have an argument that is valid or he can have an argument
with true premisses. He can't have both, at least in the same single argu­
ment, consistently with the requirements of deductive validity.
The fallaciousness of equivocation in such a case has been explicated
by Woods and myself (1979) in terms of the psychological theory of cogni­
tive dissonance: whichever way the subject of the argument chooses he is
faced with inconsistency. Here, he must choose between invalidity and
unsoundness. He can only have validity at the cost of false premisses. Or,
alternatively, he can only have true premisses at the cost of conceding the
invalidity of his argument. However, the way (being offered by the sender
of the argument) to resolve the dissonance is to amalgamate the two argu­
ments into one pseudo-argument having the appearance of both soundness
244 INFORMAL FALLACIES

and validity. Presumably, then, this move is the one hoped for by the one
who proffers the arguments so combined. Hence the justification for think­
ing of equivocation as a fallacious move to make in argument as a dialogue.
But what makes the fallacy work is the contextual shift between the
two premisses. The first premiss is only plausible in one sense of 'end,' the
second premiss only plausible in the other sense. It is precisely this duality
of context that produces the tug towards invalidity. It seems then that the
fallacy is connected to a duality between the need for true (or plausible)
premisses, and the need for a valid argument.
This duality becomes significant in the context of games of dialogue
because a primary purpose of dialogue is for the one participant to convince
or persuade the other that a certain proposition is true. In logical games of
dialogue, the object of the game is for the respondent to take as premisses
propositions that are commitments of the opponent, and then deduce the
proposition at issue, using the rules of inference only, from this set of pre­
misses exclusively. Having done this, the respondent wins the game. But to
do this, a player needs two things — both a valid argument, and a set of
premisses which his opponent will concede. Hence the duality that sponsors
equivocation is indeed present in logical games of dialogue.
From the point of view of the respondent who wishes to deploy equivo­
cation as an argument against his opponent, the problem in our example is
as follows. His conclusion to be proven is the (false) proposition 'Death is
the perfection of life.' If he disambiguates 'end' consistently as 'goal' in
both premisses he gets argument (4). If he disambiguates 'end' consistently
as 'perfection' in both premissses he gets argument (5). Both are valid (in
classical logic), but each has one untrue premiss. To get both premisses
true, he needs to disambiguate one premiss one way, one premiss another
way, as in (6). But of course he can't do this, for (6) is invalid. So what he
does is to offer (3) to his opponent, hoping or expecting that the opponent
will disambiguate (3) as (6). Yet (3) has a form that could appear valid in
the sense that (3) could potentially be disambiguated as (4) or (5), both
valid arguments. What the opponent is really pushing is the invalid (6), but
under the guise that it could be valid as (4) or (5).
Is there some way of stopping this sort of move? One way is for the
opponent to require that each term be disambiguated consistently in the
way throughout the entire argument. Then we get only valid arguments like
(4) and (5). If the opponent can do that successfully, he can always combat
the respondent's attempts at equivocating successfully. But it seems to be a
 EQUIVOCATION 245

presumption of the workability of the fallacy of equivocation in realistic

argumentation that the opponent may often not be able to do that. In
superficial equivocations like (3), he can do that. But Hamblin discusses
cases of "deep" equivocation where disambiguation may be problematic.
Under these circumstances, is there some other way to combat equivo­
cation? Perhaps there is another way. To make equivocation work, the
respondent needs a set of premisses true on some disambiguation — like
those of (6) perhaps, which may not be consistently disambiguated. He
needs these premisses to seem to imply a conclusion that may be unambigu­
ously false. Is there a set of inference-rules that would never allow such
implications? Such a set, could, it seems, stop equivocation.
To see what is needed to stop equivocation, we need to look at the pur­
pose of the equivocation in making his equivocation in a game of dialogue.
The potential equivocator as arguer needs to prove a conclusion .  may
be his thesis to be proven as set by the game, or it may be some interim con­
clusion he needs to prove his ultimate thesis. Suppose he can find no suita­
ble set of premisses his opponent will accept that entail C. As an alternative
strategy, he may search for some set of sentences that can be disambiguated
in at least two ways so that at least one disambiguation is true and one is
false. His initial difficulty was posed by the problem that the true disam­
biguation does not entail C. But embarking on a strategy of equivocation,
it could be enough if the false disambiguation implies C. For if the oppo­
nent is unwary, he may accept the unambiguated set of premisses as both
true and C-entailing.
How can the opponent stop this? He can disambiguate the premisses.
But suppose he can't do this, or at any rate, can't be sure he has done it. Is
there some way he can screen out potentially equivocal arguments? There
is no need or purpose in screening out and rejecting the true propositions
that are disambiguations of any sets of sentences offered by an arguer.
What the opponent needs is to screen out the false ones that entail  by
valid implications. Normally the opponent is well-advised not to accept
false (implausible) propositions anyway, for that is in the nature of his strat­
egy in the game. But he needs to be on his guard to not unwittingly accept
false propositions that imply C, his adversary's conclusion to be proved.

2. Vagueness and Criticisms of Equivocality

In an equivocal argument, we have the same term that occurs twice in
the argument — once in one proposition and once in another. To make
246 INFORMAL FALLACIES

both propositions come out plausible, we are tugged to interpret the same
term in two different ways. It is the plausibility of the proposition that pulls
the equivocal interpretation along, so to speak.
But very much the same sort of phenomenon can occur where the terms
that occur in the argument are vague rather than ambiguous. In such an
argument, we may have the same term that occurs twice in the argument —
once in one proposition, once in another. To make both propositions plaus­
ible, we are tugged to interpret strictly in the one occurrence, loosely in the
other. Yet it may be that if the degrees of strictness and looseness were
traded around, one or both propositions would come out as implausible.
A good example of this sort is given by Cederblom and Paulsen (1982,
p.59).
1. Getting married involves promising to live with a person for the
rest of one's life.
2. No person can safely predict compatibility with another person
for life.
3. If two people aren't compatible, then they can't live together.
4. No one should make a promise unless she or he can safely predict
that she or he can keep it.

Therefore, no one should get married.

If you look at each of the premisses of this argument, it seems possible to
interpret them as being fairly plausible. The terms that occur in these prop­
ositions, like 'compatible' and 'safely predict' are, after all, vague enough
to allow you to interpret each premiss — each premiss taken individually,
that is — as coming out as admissibly true, subject to charitable interpreta­
tion of its terms. However, if you put all four premisses together, they
imply a conclusion that you may not feel is plausible, or admissibly true. So
the person to whom the argument is directed may feel the same kind of dis­
sonance as in an equivocation. The premisses are individually plausible.
One wants to accept them. But collectively they imply an implausible con­
clusion. So can one interpret the premisses as true?
Looking more closely at the premisses, we begin to see some incongru-
encies. Take the word 'compatible.' We could interpret this term loosely,
meaning that two individuals are compatible unless they have severe disag­
reements and antipathies all the time whenever they are in contact. Under
this interpretation, premiss 3. certainly comes out as highly plausible. But
 EQUIVOCATION 247

by the same interpretation, premiss 2. seems to come out as much less

plausible. For surely it is fairly safe to predict that two individuals will be
compatible if genuine incompatibility requires that they have severe disag­
reements and antipathies all the time, whenever they are together.
On the other hand, we could interpret the term 'compatible' more
strictly. We could mean that two individuals are compatible only if they get
along very well and smoothly in their relationship with each other. By this
strict interpretation, even minor disagreements or antipathies would count
as enough to determine a noncompatibility. 1 By this interpretation, fewer
couples would count as "compatible."
Now notice what happens to the plausibility of premisses 2. and 3.
when 'compatible' is interpreted strictly. By this strict interpretation, even
if two people are not compatible, it could still be quite possible that they
could get along and live together if they put some effort into the relation­
ship. On this interpretation, lots of couples count as "incompatible," so it
might be reasonable to assume that some of them could still live together
despite their "incompatibility." On the strict interpretation, premiss 3.
becomes not very plausible, even if premiss 2. is much more plausible. Pre­
miss 2. is more plausible on the strict interpretation than on the looser
interpretation. For on the strict interpretation, fewer couples count as com­
patible — even minor disagreements count as non-compatibility. But by
this criterion, it is of course much easier to safely predict non-compatibility
of two persons over a prolonged period.
In short, vagueness can operate, in an argument, in much the same
way that ambiguity operates to sponsor equivocation. The vague term may
allow enough latitude for strictness or looseness of interpretation so that
two premisses can each seem plausible, each interpreted separately from
the other. The error emerges when we try to combine validity with the
plausibility of the individual propositions that make up the argument. To
make the individual premisses come out true, or at least plausibly true, a
heterogeneous interpretation of strictness versus looseness must be
imposed on some pair of terms in separate premisses. But to make the argu­
ment come out valid, a homogeneous interpretation must be given to the
terms of the argument. By trying, impossibly, to have it both ways, we
equivocate, or commit the fallacy of vagueness analogous to equivocation
which is illustrated by this example.
With reference to this particular example, other criticisms along the
same lines can be advanced in connection with some remaining terms.
248 INFORMAL FALLACIES

Cederblom and Paulsen (1982, p.60) add that the expression "safely pre­
dict" admits of an ambivalence of interpretation as well. What percentage
of accuracy should count as a "safe prediction?" A high standard that would
tend to make premiss 2. true would make premiss 4. tend to come out false.
For sometimes promises can't be kept for reasons that perhaps could not
have been reasonably foreseen by the person who made the promise. But
contrarily, a more relaxed standard of what counts as a safe prediction
would make premiss 4. begin to seem much more plausible only at the cost
of making premiss 2. much less plausible. The same tug-of-war between
validity and plausibility of premisses is evident in the variability of "safely
predict" as we found in "compatible."
Terms in natural language are virtually always vague to some extent.
Therefore this phenomenon of the equivocability of arguments with vague
terms in them is a constant danger in evaluating arguments in natural lan­
guage. What the phenomenon suggests is that there is always a certain
potential latitude open to a critic in evaluating and interpreting an argu­
ment in natural language. The critic must — if he or she is fair — try to
interpret the premisses and conclusion, if there is scope for strictness or
looseness of interpretation, to make the propositions in the argument come
out as being plausible. However, there are limits in how far a critic can or
should go. One such limit is posed by equivocation. The sentences in the
argument should charitably be interpreted to make each of them most
plausible, but not at the cost of equivocating. That is, the concessions to
chanty can only go so far when a choice must be made between plausible
premisses (or conclusion) and a valid argument. When the critic determines
such a clash, he should propose to the arguer that a choice must be made.
He should, in other words, advance the criticism of equivocation, request­
ing that the proponent of the argument make a choice one way or the other.
Cederblom and Paulsen show keen insight in suggesting that this sort
of case reveals that the evaluation of the argument is fundamentally a
dialogue process. The critic must ask the arguer what he means by certain
terms. The arguer must then respond by clarifying his meaning. Hamblin
(1970), I feel, has made a kindred deep observation in noting that equivoca­
tion as a fallacy is best seen as a "point of order" or procedural clarification
in dialogue between two participants in dialogue.
Another important lesson of this case is that there is a fundamental
connection between equivocation (including the kind of equivocation
associated with vagueness) and enthymemes. Evaluating equivocal argu-
 EQUIVOCATION 249

ments, and perhaps even all arguments in natural language, is a matter of

making certain fair or charitable presumptions about what the arguer meant
to say or include, given the benefit of the doubt. But this is the very same
problem posed by cases of enthymemes. If the argument is a written one, or
if the arguer is not present to defend or clarify his argument by further
dialogue, criticism of the argument can only proceed on the basis of "fair"
presumptions about what the arguer would have, or should have, added or
further specified. Hence all criticisms by way of equivocation or
enthymemes are in an important sense provisional. They are procedural
clarifications that invite the arguer's response. They are therefore dialogue-
relative, and are better seen as criticisms inviting replies than as conclusive
refutations. But then such dialogue-relativity has emerged as characteristic
of virtually all the traditional fallacies. Hence a "fallacy" may be better seen
more as an objection that can possibly be met than as a refutation that
shows, once and for all, that an argument is incorrect.

3. The Problem of Subtle Equivocations

Hamblin (1970) makes a distinction between gross and subtle equivo­
cations. Most cases cited as fallacies of equivocation in the traditional logic
texts are either not arguments, and hence are more correctly described as
ambiguities rather than equivocations, or are simplistic cases that would be
unlikely to be much trouble to a serious reasoner. For example, Hamblin
(1970, p.14) notes that at least three of the textbooks he consulted gave the
example: "Some dogs have fuzzy ears; my dog has fuzzy ears; therefore my
dog is some dog!" This sort of example is not very helpful however, for as
Hamblin rightly points out, "... we are hardly capable of being deceived by
any serious chain of reasoning exploiting the double-meanings in the state­
ments about them" (p.15). Hence a good case of the fallacy of equivoca­
tion, worth our serious study, should be (1) an incorrect (invalid) argu­
ment, (2) based on meaning-shift, and (3) the putting forward of which is
part of a strategy of deception or significant mischief in argumentation.
Are there such arguments? Hamblin goes on to give an example of a
subtle equivocation that might fit the bill.
(8) All acts prescribed by law are obligatory.
Non-performance of an obligatory act is to be disapproved.
Therefore, non-performance of an act prescribed by law is to be
disapproved.
250 INFORMAL FALLACIES

This case is an argument, at least so I think, that could indeed be subtle

enough to cause a good deal of trouble in real arguments. And there could
be little doubt that in fact it has caused a lot of trouble in many an argument
on ethics.
But having gotten a realistic example, Hamblin is immediately led to
recognize a general problem posed by it. How could it truly be a fallacy, if
in fact it is subtle enough to truly deceive an arguer? To see the problem,
we have to start with Hamblin's basic assumptions about equivocation as a
fallacy.
First, Hamblin requires that in order to be a fallacy of equivocation, an
argument must (a) be an invalid argument based on meaning-shift, and (b)
"... we must assume that the perpetrators of the argument either deceive
themselves, or set out to deceive other people, into thinking the argument
valid" (p.292). But if we consider the argument (8) above, we can see that
someone could be deeply deceived by it. That is, someone who advanced
this argument might be so convinced by it that he identifies the class of mor­
ally obligatory acts and the class of acts prescribed by law as being perfectly
equivalent in meaning. We can, Hamblin suggests, suppose someone to be
so deeply deceived by such an argument that it creates for him a special pat­
tern of use of the words involved in it. It could be as if this person is
stipulating that, for the purposes of his argument, 'morally obligatory acts'
and 'acts prescribed by law' shall mean the same thing. But if we can grant
such an assumption then there is a serious problem with equivocation as a
fallacy — for the arguer who is deeply "deceived," the argument he
advances is no longer meant by him to be equivocal. For him there is no
double meaning, and hence no shift of meaning. To any accusation of
equivocation, he can or should reply, "According to the meaning of these
words that I intend to advance as part of my argument for my position,
these phrases mean the same thing. There is no double meaning, and there­
fore no equivocation." How is the critic to reply? If stipulative meaning is
sometimes possible in argument, as surely it should be, then the critic
appears to have no foolproof reply. Hence Hamblin's problem — how can
you ever make a charge of a fallacy of equivocation stick, if the case at issue
may be truly subtle enough to fool somebody?
Part of the problem is unravelling the meaning of 'fallacy.' I think we
need to question Hamblin's assumption — as part of the way he formulates
the problem — that for there to truly be a fallacy of equivocation, some­
body must be deceived. Generally, we have seen that there is a question-
 EQUIVOCATION 251

able psychologism implicit in the claim that a fallacy must be an invalid

argument that seems valid to someone. And in this instance, I would like to
resist the idea that we must assume that the perpetrator of a genuine fallacy
of equivocation either deceives himself or those to whom his argument is
directed. Instead, access to equivocation as a form of fallacy open to
reasonable analysis should be sought in the dialectical strategy of the arguer
to prove something to another participant in the logical game of dialogue
both are engaged in. If equivocation is a fallacy, it is because it is a strategy
of dialogue that may enable an arguer to prove his conclusion, by the rules
of the game, from his opponent's commitments.
Hamblin's deeper formulation of the problem goes on to recognize and
bring out the dialectical nature of the problem of equivocation as a fallacy.
The underlying problem is to see how "deception" or "fault" can have a
dialectical analysis in the context of equivocation. Here we no longer mean
"deception" in a psychological sense, but refer to normative rules or proce­
dures of rational dialogue, and violations thereof.
Hambln recognizes, however, that the basic structure of equivocation
as a fallacy has to do with a kind of shift that takes place between the arguer
and the one to whom the argument is directed. Controlling this characteris­
tic shift has something to do with procedural rules of dialogue.
First, Hamblin (p.293) notes that equivocation has something to do
with the dialectical theory of truth and falsity. For the proof that an argu­
ment is equivocal seems to require that the argument is, in some sense,
invalid. To make a charge of equivocation supportable, a critic has to show
that the premisses of the argument may be taken to be true while the con­
clusion may be taken to be false. What I would like to think Hamblin is
moving towards, in these observations, is the position that equivocation is
essentially tied to shifts in the burden of proof in moves of play in games of
dialogue, and also to an arguer's evaluations of the plausibility of the pre­
misses and conclusion of an argument. If for 'truth' we read 'plausibility'
and for 'falsity' we read 'implausibility' the character of equivocation as a
strategy of proving in logical games of dialogue can emerge. However, such
a suggestion arises naturally within the context of the present theory that
plausibility is what corresponds in dialectic to truth. The conclusion to be
drawn then is that equivocation as a fallacy has to do with shifts of plausibil­
ity between moves in a game of dialogue.
Some of the problems posed by Hamblin can be solved within the
theory of the fallacy of equivocation as a strategy of plausibility — disso-
252 INFORMAL FALLACIES

nance between players in a logical game of dialogue. Let us go back to the

subtle example of equivocation (8) given by Hamblin. Is this argument a
fallacy of equivocation? To decide whether it is or not, we do not need to
worry whether in general it "seems valid" but is not. Instead, we should
look to the question of how it might be deployed in a game of dialogue.
Characteristically, (8) is a fallacy of equivocation if put forward by a
participant in dialogue where his opponent is tugged to interpret some term
in one proposition in one way, in another proposition another way, in order
to make both propositions come out as plausible to that opponent. The
other elements of equivocation characteristically are that under such a dis­
ambiguation, the argument is invalid, and that under some other admissible
disambiguation the argument is valid. In a quite natural context, (8) could
easily be portrayed as just this very sort of argument bundle. It is really a
multiplicity of arguments rolled up into one package. And as Hamblin
rightly proposes, it is subtle enough to deeply deceive participants in a
realistic sort of context of argumentation which could be constructed as its
dialectical setting. Hence it is reasonable to grant that (8) is quite a useful
example of equivocation.
But why is (8) a fallacy of equivocation? It is a fallacy precisely to the
extent that a reasonable assignment of plausibility-values to the premisses
makes them require different disambiguations of obligatory' to be plausi­
ble to one to whom the argument is addressed. Moreover, under such a dis­
ambiguation, the argument is invalid. Furthermore, the argument has at
least one valid disambiguation (with at least one implausible premiss).
So Hamblin is right that invalidity and a meaning-shift are characteris­
tically involved. The shift has to do with the dialectical "truth" or "falsity"
of the premisses, because these notions become translated into "plausibil­
ity" and "implausibility" in the context of strategies of dialogue.
But one of Hamblin's most profound problems posed by (8) still
remains. What if the proponent of (8) is in reality deeply deceived himself,
in the sense that for his personal or emerging usage, the two disambigua­
tions of 'obligatory' are in fact equivalent or synonymous? However its reci­
pient takes it, can (8) still qualify as a fallacy if it contains no propositions
that are truly ambiguous for its sender?
In other words, what's really fallacious in regard to (8)? Is it the set of
three propositions as argument that are fallacious? Is it the proponent of
the argument who commits the fallacy? Or is it the recipient of the argu­
ment who may commit a fallacy of equivocation, even if the proponent sin-
 EQUIVOCATION 253

cerely propounds his own consistent or non-ambiguous usage? The diffi­

culty here is compounded by reasonable doubt on how to define 'argu­
ment.' Is the argument a set of sentences, or a set of propositions, or are
the arguers included as well? Or are their moves or commitments somehow
included as part of the "argument?" These problems challenge our concep­
tions of fallacy and argument.
The best answer is that to understand (8) as a fallacy of equivocation
we have to look at it as a possible move of argument in a game of dialogue.
Presumably, the context is that a proponent is advancing (8) with the
strategic objective of getting his opponent to accept the conclusion. (8) only
makes sense as a fallacy if that opponent finds both premisses plausible only
at the cost of interpreting 'obligatory' one way in the one premiss, the other
way in the remaining premiss. Whether in fact that proponent finds the pre­
misses ambiguous or not in his own semantical lexicon is not at issue. The
question of whether there is a fallacy of equivocation or not should turn on
whether that proponent advances (8) and whether the opponent interprets
the premisses in the only plausible way as being ambiguous. Whether (8) is
fallacious or not is a question of the strategy adopted by the proponent in
using (8) as a move as part of the game of dialogue at issue. These consider­
ations partly depend on the recipient, on whether he finds the premisses
plausible or not. But they depend greatly on the proponent, who constructs
the argument using a strategy based on the assumption that the recipient
will disambiguate on the basis of what he takes to be plausible. The fallacy
involves aspects of both parties.
In other words, the term 'argument' needs to be construed broadly
enough to encompass the propositions that make up the premisses and con­
clusion, and also the dialectical context. The dialectical context includes the
description of the theses to be proven by the participants, the participants
themselves as players of the game, the rules of the game including the
requirements of validity, and some account of the positions of each player
so that there is a basis for evaluating some propositions each will take as ini­
tially plausible. Given all this, equivocation can emerge as a fallacy. It is the
fallacy of one player putting forth several arguments and therewith imple­
menting the strategy that his opponent will accept the conclusion by incor­
rectly taking plausible premisses from one disambiguation and a valid argu­
ment from another disambiguation. The fallacy is best analyzed as a charac­
teristic type of strategy in a game of dialogue.
These remarks may be helpful in deepening Hamblin's insights, but
254 INFORMAL FALLACIES

they still do not solve all the problems posed in his chapter on equivocation.
It could still be possible for a disingenuous equivocator to claim that, by his
criteria of meaning, there is truly no ambiguity and therefore no equivoca­
tion. As is usual with the fallacies, the problem of really "nailing down" the
allegation of fallacy in realistic dialogues is non-trivial.

4. Deep Deception and Equivocal Dialogue

Given our broadening of the concept of argument to include the basic
elements of regulated dialogue between two arguers, let us return to
Hamblin's central problem with equivocation as a fallacy. Consider the case
of an arguer, White, who advances argument (8), but is so deeply deceived
by it that it is fair to say of him that he makes no distinction between the
class of acts required by law and the class of morally obligatory acts. In
White's lexicon, these two classes are semantically equivalent. Let's sup­
pose as well that the other participant in the dialogue, Black, is confronted
by White, who advances (8) at some point in the dialogue. Moreover, let's
assume that the objective of the game of dialogue is for each player to
prove his thesis to the other. What should be said of White's deployment of
(8) in these conditions?
Let's suppose that Black, at the next move in the dialogue, replies to
White's move as follows: "Your argument is an equivocation. The term
'obligatory act' is ambiguous." And further let's suppose that White replies
to Black's criticism in the defensive way Hamblin suggests as being possi­
ble: "According to the meaning of the words I intend to advance in my
argument, this phrase is not ambiguous. There is no double meaning, and
therefore no equivocation." How is Black to proceed? Now if Hamblin is
right that White's reply is possible, and could be legitimate, the problem is
this — how could Black ever hope to make his charge of equivocation stick?
In the context of dialogue given above, the problem can be amelior­
ated to some extent by the following observation. Once White concedes
that he means both 'legally required' and 'morally required' by the term
'obligatory act' in both of his premisses, Black has no further need to make
the charge of equivocation stick. Black should forget the charge of equivo­
cation at this point and reply as follows: "Under that univocal (in your
semantics) interpretation of 'obligatory act' your argument is valid, I con­
cede. However, under that same interpretation, neither premiss is plausi­
ble. For only a legal positivism of the most rigid and extreme sort will
accept the proposition that non-performance of every act prescribed by law
 EQUIVOCATION 255

is to be morally disapproved. Most of us justifiably think that when a law is

foolish or immoral, there may arguably be exceptions to viewing non-per­
formance of it as subject to moral disapproval. Likewise, only an extreme
legal positivist would accept the proposition that all acts prescribed by law
are morally obligatory. Hence your argument is valid, but I dismiss it
because both of the premisses, as you intend them, are unproven and unac­
ceptable." In short, Black does not need to make his charge of equivocation
"stick" once White reacted to it by opting for his univocal interpretation of
the premisses. Once White so reacted, the charge has done its work, and
Black can proceed to attack the weak premisses. For the charge of equivo­
cation has done its job of exposing that weakness to rebuttal or criticism.
We can see here that plausibility plays an important role in equivoca­
tion as a form of argument and reply in dialogue. The argument was only
plausible when the premisses were taken ambiguously by Black. Once the
ambiguity is perceived, the argument ceases to be plausible for Black.
White may have revealed himself as an extreme legal positivist, but one
may well assume that as part of a realistic profile of the dispute, Black may
not be committed to any like positivistic presumptions. On such a profile,
Black's reply should be clear and pressing.
Once Black has mounted the charge of equivocation and White has
acknowledged that for him, moral obligation and legal obligation are the
same thing, the argument can be re-written as follows.
(9) All acts prescribed by law are both morally and legally obligato­
ry.
Non-performance of an act that is either morally or legally oblig­
atory is to be disapproved.
Therefore, non-performance of an act prescribed by law is to be
disapproved.
White's argument, so restated, is non-equivocal for both Black and White.
Perhaps White's interpretation of his argument could be made even clearer
by re-stating the premisses after the following fashion: all acts prescribed by
law are legally obligatory, or, to put it in an equivalent way, morally oblig­
atory. This way of putting the premiss brings out that White has clearly
opted for a strong form of legal positivism in advancing (9) the way he does.
To gain a more realistic appreciation of whether any "fallacy" is com­
mitted in a case like this one it is necessary to fill in some of the dialectical
context of (9). Let us suppose that White has taken up the position of
256 INFORMAL FALLACIES

extreme legal positivism. For White, moral obligation is perfectly equiva­

lent to the requirement of law in some given jurisdiction. In relation to
White's position, (9) as advanced by White is — at least for White — non-
equivocal. But must it follow that White has committed no fallacy of
equivocation in advancing (9) as an argument in dialogue with Black? Let's
suppose that Black is not committed to a position of legal positivism. Then
for Black, the second premiss is false or implausible. Could White have
committed a fallacy of equivocation then, by proposing (9) to Black?
A case could be made out for criticizing White as follows. Suppose that
the issue being disputed by White and Black is precisely that of the truth of
legal positivism. White defends it and Black rejects it. In such a dialogical
context, the conclusion of (9) could be an important thesis for White to
establish as part of his argument for legal positivism. Therefore, if the game
is a dispute, Black is committed to rejection of the proposition 'Non-perfor­
mance of [every] act prescribed by law is to be disapproved' where 'disap­
proved' means 'morally disapproved.' Once White takes the step of advanc­
ing (9), Black is immediately committed by the strategic dictates of his posi­
tion, to driving a wedge between the two meanings of 'obligatory.' For
clearly Black should be committed, in virtue of his rejection of legal
positivism, to the rejection of the following two propositions.
(10) All acts prescribed by law are morally obligatory.
(11) Non-performance of [every] legally obligatory act is to be [mor­
ally] disapproved.
These two propositions, taken together, could in effect define the basic
position of strong legal positivism that an act is legally binding if, and only
if, it is morally good.
By getting White to concede that in White's pattern of usage there is
no ambiguity, Black can make it clear why he, Black, should not be com­
mitted to either premiss of (9), as White interprets them. So while it is true,
perhaps, that Black may not be able to make his charge of equivocation
stick, that is not a serious problem for Black if he follows up the charge of
equivocation in the appropriate way.
In short then, once Black recognizes the ambiguity and makes it clear
by his charge of equivocation, the dialectical sting is taken out of White's
argument. White's insistence that the argument is not an equivocation,
from his own point of view, does not make it an acceptable argument from
Black's point of view. And in logical games of dialogue where the objective
 EQUIVOCATION 257

is to prove your own conclusion from your opponent's concessions, it is

Black's point of view that is the more significant in connection with ruling
on fallacies of equivocation. The argument may be univocal from White's
position. But once seen clearly as a univocal argument by Black, the argu­
ment is no threat to Black's position.
It is only when the ambiguity is not recognized by Black, as a signifi­
cant ambiguity in relation to his position, that the equivocal argument could
be a real threat. To make the argument plausible in relation to his position,
Black could be tempted to interpret one premiss one way and the other pre­
miss the other way. Yet if, in so doing, he also accepts the validity of the
argument, he has been duped by the equivocation. For to accept the plausi­
bility of the premisses and the validity of the argument at the same time is
to erroneously accept a bundle of faulty arguments as if it were one valid
argument with plausible premisses. And that is precisely the fallacy of
equivocation.
So we see that equivocation is a dialogical fallacy. One party, White,
advances an argument that may be, for him, non-equivocal. But clearly the
"argument" is an equivocation if, in relation to its receiver's position, one
premiss is only plausible if taken one way, the other premiss only plausible
if taken another way, and neither way yields a valid argument. Such an "ar­
gument" is not an argument at all, in relation to the recipient's position. It
is many arguments, even if it may be taken in relation to another position
as one argument. As one argument, however, it may not have plausible
premisses in relation to an arguer's position.
Who commits a fallacy of equivocation then, in the case of (8)? Is it
White or Black? I think we should rule that it is White who commits the fal­
lacy. It is White who needs to convince Black to accept the conclusion of
(8). Yet the premisses offered by White are most plausibly interpreted
equivocally by a non-positivist like Black. However, if Black were to accept
the "argument" and its conclusion based on his equivocal interpretation of
it, it is fair to rule that he too has participated in the fallacy.
What does "fallacy" mean then, in the context of equivocation? Here,
I think, is where we need to question our rule that a fallacy is a fallacious
argument. For an equivocation is not an argument, strictly speaking. It is a
set of sentences comprising many arguments, masquerading as one argu­
ment. Clearly then, argument is involved. But the problem is that there are
too many arguments being taken, confusingly, as one argument by the
arguer who falls victim to equivocation.
258 INFORMAL FALLACIES

To get an adequate grasp of equivocation as a fallacy, we have to study

this kind of multiplicity of move in argument in the context of strategy in a
game of dialogue. Each participant in the game has a set of commitments
that defines his position at a given point in the dialogue. The position of an
arguer may not be fully revealed to his opponent, but usually that opponent
has some rough idea of what the other player's commitments are likely to
be on the issue of the disputation. He knows that certain propositions are
likely to be thought plausible or implausible by that other arguer. Such indi­
cations of plausibility may predictably incline an arguer to interpret an
ambiguous sentence in one way or another. Where ambiguity is possible
then, equivocation becomes a possible short-cut strategem in argument. We
can see now clearly enough why it works, and why it is a fallacy.
The dialectical approach we have taken does, to a certain extent,
resolve Hamblin's problem of the disingenuous equivocator who claims his
own pattern of private or stipulative usage. For although he may be entitled
to such usage, as required by his own position in the argument, he may not
be entitled to force it on the recipient of his argument, who may have a dif­
ferent position to defend. From the latter viewpoint, the argument may be,
if not equivocal, then simply uncompelling in virtue of its unacceptable pre­
misses. At any rate, from this perspective on argument, Hamblin's problem
is not quite as bad as it may have initially seemed to be.
Still, no procedural requirement of dialogue we have adopted, so far,
rules out the equivocator's move of advancing a bundle of arguments
designed to capitalize on ambiguity and plausibility and thereby sophisti-
cally gain a recipient's acceptance of the equivocator's conclusion. The
equivocator can certainly try this in a logical dialogue-game, and it is then
a matter of "the buyer beware." It is up to the recipient to detect the
equivocation if he can.
If equivocation is possible in games of dialogue then, how can we stop
it? That is, could there be some grid or screen that a defender could use to
protect his side of the argument against the equivocator? That would seem
to be the next question of interest.
One might reflect here that the problem is posed by the assumption
that argumentation takes place in natural language, or at any rate in some
language that tolerates ambiguity. If the game of dialogue is conducted exc­
lusively in some formal language like first-order logic, no fallacy of equivo­
cation could ever be mounted. For in this sort of game, every constant is
defined univocally, and there can be no ambiguity. That model of argument
 EQUIVOCATION 259

is fine as far as it goes. But of course the problem is that if we want to

model realistic debates or disputations, we need to study arguments that
basically take place in the medium of natural language, even if some "logi­
cal" rules of procedure may be enforced, as governing some kinds of moves
and inferences. Equivocation then is truly a problem of applied or informal
logic. From a point of view of pure logical theory with no pretentions to
apply to realistic argumentation, equivocation is no problem. Where there
is no ambiguity, there can be no equivocaton.
Woods and Walton (1979) takes an approach of considering languages
that are partially regimented so that equivocation can be contended with.
However, ultimately that is a piecemeal solution. For as soon as you admit
ambiguity of any sort into the language, equivocation is possible. True, the
more the semantics of the language is regimented or formalized, the better
is the armory of the defender equipped to contend with equivocation
attacks. But unless natural language is given up entirely, the possibility of
equivocation must remain.
In realistic argumentation then, it seems that it must remain up to the
defender to be on his guard against equivocal arguments. We do not know
of any foolproof way to screen against it.
There is one sort of approach that could be tried however. Instead of
trying to construct an outer defence system in the procedural rules of logical
dialogue that would equip the defender to exclude or refute equivocal argu­
ments, could we try to modify the internal logical propositional structure of
dialogue to bar equivocal arguments of the potentially harmful sort. That
is, could we restructure the logic of 'valid argument' to bar the equivocal
argument? Proposals that look to be of this sort have recently been
advanced by David Lewis. Let us look to them.

5. Many-Valued Logic for Equivocators

The problem of equivocation in argument is that the equivocator pre­
sents more than one argument. In effect, he advances a bundle of argu­
ments at one move. Before the defender against the fallacy of equivocation
can cope with this sort of multiple move, he must sort out the different
propositions that are contained in the bundle. Once he sorts them out into
several distinct arguments, he can then react in the usual way appropriate
to logical dialogue. He can evaluate each individual argument on its merits.
Classical propositional logic, so we have argued, is an appropriate semanti­
cal structure for this evaluation in many contexts of dialogue.
260 INFORMAL FALLACIES

But there is one serious problem. In order to proceed in this way to

combat equivocation, the defender needs to recognize the ambiguous sen­
tences as ambiguous, and then sort them into the unambiguous sets of prop­
ositions he needs to make up arguments. The worst danger of equivocation
however, is that an arguer may not be able to take the first step of recogniz­
ing the ambiguity and clearly sorting out into non-ambiguous propositions.
Natural language is rife with ambiguity and vagueness that may go unde­
tected.
We could deal with this problem by having a propositional logic for
ambiguous sentences. Accordingly, a sentence could be true on some dis­
ambiguations (true-osd) or false on some disambiguations (false-osd). Once
we allow this way of evaluating sentences, there will also be some mixed
sentences. That is, some sentences will be true on some disambiguations
and false on others.
According to this way of proceeding, truth and falsity are still the prop­
erties of propositions. But of course, one sentence, if it is ambiguous, may
express many propositions. Hence a given sentence could be true on some
disambiguations and false on others. Hence, at the level of sentences, we
should say that a sentence is true-osd or false-osd. These evaluations are
the counterparts of the propositional properties of truth and falsity.
What does 'validity' mean at the level of sentences? One thing it could
mean is that in a valid argument, if the premisses are true-osd, then the
conclusion is also true-osd. In other words, in a valid argument in this
sense, you can never go from premisses that are true on some disambigua­
tion to a conclusion that is false on every disambiguation. Another possibil­
ity is the following conception of validity: if the premisses are true on all
disambiguations, then the conclusion must also be true on all disambigua­
tions.
Adopting one or both of these conceptions of validity at the sentence
level would have the decisive advantage of providing a defensive screen for
the defender against equivocation. He does not need to disambiguate the
potentially ambiguous sentences in which an argument is expressed prior to
his evaluation of the argument. Therefore, these new conceptions of valid­
ity are worth pursuing in any study of equivocation as a fallacy.
But can either of these conceptions of validity yield a propositional cal­
culus with well-defined properties? David Lewis (1982) has shown how
each of them yields a many-valued relevance logic. Let us examine both of
these logics.
 EQUIVOCATION 261

6. Priest's System LP
Validity of arguments in Priest's system LP means never going from
premisses that are true only or mixed2 to a conclusion that is false only.
This criterion of validity is what is stated by Priest's definition of the deduci­
bility relation,  Let ρ and q be propositions, and v be an evaluation of the
propositional variables (i.e. a function that takes the variables to values T,
M, or F). What Priest calls v is the extension of v to the complex sen­
tences of the language, A, B, C, ..., made up from the propositions using
the truth tables. Then where Σ is a set of sentences of the language LP,
deducibility is defined as follows (Priest, 1979, p.228).

What does this mean? It means that a sentence A is deducible from a set of
sentences Σ where you never go from Σ that is true only or mixed to A that
is false only. In other words, if the premisses have the value Τ or the value
M, then if the conclusion is deducible from those premisses, the conclusion
cannot have the value F. To put it yet another way, deductibility in Priest's
system LP is defined as the set of implications that preserves the values Τ
and M. A valid argument will never take you from one of these values to
the value F.
According to Lewis (1982, p.439), the implications that preserve truth-
osd are given by Priest's LP. What Lewis means is that he is interpreting
Priest's three values T, F, and M as applying to ambiguous sentences. There
are just these three possibilities: a sentence can be true on all its disambigu­
ations (T), false on all its disambiguations (F), or true on some and false on
others (M). Hence Priest's system LP has a concept of implication that pre­
serves truth-osd (truth on some disambiguation). This means that if a sen­
tence A implies a sentence  in LP, then if A is true on all its disambigua­
tions or true on some disambiguations and false on others, then so must 
have this same property, i.e.  must be true on all its disambiguations or
true on some disambiguations and false on others. In other words, truth-
osd is preserved by implication. You can never go from premisses that are
true on some disambiguations, by valid argument, to a conclusion that is
false on all its disambiguations.
There is one qualification yet, however. According to Lewis (1982,
p.439), "[t]he implications that preserve truth-osd are those given by the
first-degree fragment of Priest's LP." What is the first-degree fragment of a
262 INFORMAL FALLACIES

system? According to Anderson and Belnap (1975, p.151) "degree" refers

to the nesting of arrows → (conditionals). A zero-degree formula contains
no arrows at all. A first-degree formula is a formula A →  with both A
and  zero-degree (purely truth-functional) formulas.3
According to Lewis (1982), a sentence can be disambiguated three dif­
ferent ways. It can be true only, on some disambiguation. It can be false
only, on some disambiguation. Or it can be true and false, on some disam­
biguation. Take the sentence 'Scrooge went to the bank,' presuming that it
can be disambiguated two ways: (1) Scrooge went to the (savings) bank,
and (2) Scrooge went to the (river) bank. Suppose Scrooge went both to the
savings bank and the riverbank, for example, if the savings bank were on
the river. Then the sentence 'Scrooge went to the bank' is true only, on the
given disambiguation. But suppose Scrooge did not go to either the savings
bank or the riverbank. Then the sentence 'Scrooge went to the bank' is
false only, on the given disambiguation. Finally, suppose Scrooge went to
the savings bank, but did not go near the riverbank. Then the sentence
'Scrooge went to the bank' is both true and false, on the given disambigua­
tion.
Following Lewis's interpretation of logic for equivocators, we can see
how the connectives should be defined after the tables given by Priest
(1979). Let Τ be the value "true only," F be the value "false only" and M,
or "mixed," be the value "true and false." We presume (as usual) that the
negation of a sentence is the opposite value. That is, a sentence is true
(false) on some disambiguation if and only if its negation is false (true) on
some disambiguation. Suppose then that the sentence 'Scrooge went to the
bank' is true only on some disambiguation. What shall we say of its nega­
tion, 'Scrooge did not go to the bank?' We are presuming here that, in fact,
Scrooge did not go to the savings bank and he did not go to the riverbank.
Clearly then, the sentence 'Scrooge did not go to the bank' is false only, on
the given disambiguation. Whichever way you disambiguate 'bank' the sen­
tence comes out false.
Similarly, if a sentence is false only, its negation must be true only.
But what if 'Scrooge went to the bank' is both true and false (mixed),
on a given disambiguation. Then its negation will have the opposite value.
It will be false and true, on the same disambiguation. In other words, the
negation of a mixed sentence will also have mixed values.
What we have said then can be summed up by the following table for
negation.
 EQUIVOCATION 263

A ΙΑ
Τ F
M M Table for Negation
F Τ
Similar reasoning leads to the formulation of a table for conjunction. We
need to look at the conjunctive sentence 'Scrooge went to the bank and
Don went to the plant' under different conditions of the following disam­
biguation.

Some of the rows of the table are straightforward. If A is true only and  is
true only, then    is true only. If A is true only and  is mixed, then A
  is mixed. These straightforward values are given by the eight filled
rows of the table below.
A  A ^
Τ M M
M Τ M
M M M
τ τ τ Table for Conjunction
τ F
F Τ
F F F
M F
F M
The remaining four rows of the table, left blank above, require more care­
ful consideration. Consider the third blank row from the top above. What
if A is mixed but  is false only? What value should    have?
Suppose, in our example, that Scrooge truly goes to the savings bank
but it is false that he goes to the riverbank. And suppose that it is false that
Donald goes to the factory and it is also false that he goes to the vegetation.
What are we to say of the value of 'Scrooge went to the bank and Donald
went to the plant' on the given disambiguation? Our first thought might be
to say that this sentence is mixed, since there is some truth in it as well as
falsehood.
264 INFORMAL FALLACIES

But we need to reflect on this interpretation a little more. What do we

mean by 'and?' Presumably we mean (as usual) that    should only be
true if both A and  are true. If either component is false, the whole con­
junction    must come out false. Reasoning along these lines, if the one
conjunction, say B, is false only on the given disambiguation, then the
whole conjunction,   , should be false only, It doesn't matter that the
other conjunct, A, is mixed. Hence we should say that the whole sentence
'Scrooge went to the bank and Donald went to the plant' is false only if
'Donald went to the bank' is false only, never mind whether 'Scrooge went
to the bank' is both true and false.
The same reasoning that applies to the bottom pair of blank rows in the
table for conjunction can also be applied to the top rows. Even if one con­
junct is true only, it is enough to make the whole conjunction 'false only' if
the one conjunct is false only. The presumption is again the usual one that
a conjunction is true if and only if both conjuncts are true. If one is false,
the whole conjunction is false. By analogy, if one is false only, the whole
conjunction is false only.
But there are two ways of completing the table for conjunction, each of
which could possibly be justified. The one considered above is given on the
right. But we could appreciate how the one on the left may be considered
as well.
A  A ^  _A  A ^ 
Τ M M τ M M
M τ M M τ M
M M M M M M
τ τ τ τ τ Τ
τ F M τ F F
F Τ M F Τ F
F F F F F F
M F M M F F
F M M F M F
Suppose one conjunct is false only, on some disambiguation of a conjunc­
tive sentence. Is the whole conjunction mixed, or is it false only? The
former alternative is given by the left table, the latter by the right table.
The left table says that 'A and B' is always mixed unless both A and 
are true only or false only. Then 'A and B' has the same value as its compo­
nents. According to this view of conjunction, a mixed value anywhere
 EQUIVOCATION 265

always produces a mixed value for the whole conjunction. We could say
that according to this view of conjunction, mixture is infectious.
The right table says that whenever either component has the value
'false only' then the whole conjunction has the value 'false only.' Otherwise
the conjunction is mixed whenever either component is mixed. According
to this view of conjunction, a 'false only' value anywhere always produces a
'false only' value for the whole conjunction. We could say that according to
this view, 'false only' is infectious.
The difference between the two tables only concerns the four rows
where one conjunct is false only but the other is either true only or mixed.
Should we say that the conjunction is false only, in those rows, or should we
say that it is mixed?
What should win out here, in the end, is the analogy with the 'and' of
classical logic, where a false conjunct is enough to make the whole conjunc­
tion false. Similarly here at the level of sentences, if any conjunct is false
only, then the whole conjunction should have the value 'false only' even if
other parts of it are true on some or all disambiguations. In other words,
'and' at the level of sentences should retain this much of its logic in common
with 'and' at the level of propositions.
Following these lines of reasoning, the sentential logic for equivocators
is that outlined by Priest (1979, p.228) as follows. Although the class of
tautologies in the new logic LP turn out to be the same as those of classical
propositional calculus, the deducibility relationship is changed. The follow­
ing inferences do not hold in LP.

One can see that LP is a relevance logic insofar as disjunctive syllogism and
reasoning from inconsistent premisses to any conclusion both fail. Yet it is a
somewhat unusual relevance logic, for modus ponens and modus tollens both
fail as well. However, the following principles of classical logic are preserved
in LP.
266 INFORMAL FALLACIES

Hence it seems that we have enough of classical logic left in LP to make it a

potentially interesting way of evaluating ambiguous sentences for the canon
of validity we have in mind.
To check it out further, let us turn to some examples, both to see how LP
works generally, and to see how effectively it could be applied to evaluating
charges of the fallacy of equivocation in realistic argumentation.
We conclude this section by summarizing the basic constants as defined
by tables for LP. Since 1 and  have already been defined, we could define
Α ν  as 1(1A ^ IB). And we may define A →  as 1Α ν Β. 4 But just to
be clear, we give below the tables for ν , → and ^
A  A vB A →B A ^ 
Τ M Τ M M
M Τ Τ Τ M
M M M M M
τ τ τ τ τ
τ F τ F F
F Τ τ Τ F
F F F Τ F
M F M M F
F M M Τ F
Then all we need to remember is that 1A always has the opposite value of
A (takes Τ to F or vice versa) except where A has the value M. Then 1A
also has the value M. Here we can clearly see why modus ponens fails
in LP. In row 8 above, A and A →  have the value M while  has the
value F.

7. Applying LP to the Fallacy of Equivocation

How could restricting validity to validity-osd help us to deal with fal­
lacies of equivocation? To pursue this question, consider the following
argument.
(12) If Scrooge goes to the bank on Thursday, he makes a financial
transaction on Thursday.
Scrooge goes to the bank on Thursday and does some fishing.

Therefore, Scrooge makes a financial transaction on Thursday.

This argument could, in the right disputational context, turn out to be justi-
 EQUIVOCATION 267

fiably called an instance of the fallacy of equivocation. Why? Under the

most plausible disambiguations of its terms, the premisses are both true and
the conclusion is false. The most plausible context for the first premiss
suggests that 'bank' should be disambiguated as 'savings bank.' While the
most plausible context for the second premiss suggests that 'bank' should be
disambiguated there as 'riverbank'. So disambiguated in both premisses
however, the argument comes out invalid in classical logic (and in LP, for
that matter). We have the mark of an equivocation then. Under some dis­
ambiguations you get a valid argument, but always one with at least one
implausible premiss. Under other disambiguations, you have plausible pre­
misses, but the argument is invalid.
What happens if we turn from classical validity to validity-osd (as mod­
elled by LP) as our criterion of valid argument? Clearly the argument is not
valid-osd. According to the most plausible interpretation of the premisses
— where 'bank' means 'savings bank' in the first premiss, and 'riverbank' in
the second premiss — both premisses can be true-osd without the conclu­
sion having to be true on every disambiguation. It could be that Scrooge
goes to the riverbank on Thursday, and makes no financial transaction at
all.
Reviewing this particular argument, one can see why modus ponens is
not a valid form of inference in LP. If the premisses are true on some dis­
ambiguation, it need not follow that the conclusion is true on some disam­
biguation. In this particular case, the conclusion may be unambiguously
false.
Now the problem of equivocation arose as a fallacy because an argu­
ment bundle containing ambiguous sentences could be valid on some dis­
ambiguations (at the cost of having false premisses) and at the same time
invalid on other disambiguations (with true premisses, however). Using LP
instead of classical logic, this problem no longer arises. For an argument
bundle is now either valid-osd or not, no matter how you divide it up into
arguments.
Hence the adoption of LP as the logic of sentences in a game of
dialogue does show promise of helping us to cope with the problem of
equivocation. But we need to consider further cases.
How could Lewis's interpretation of Priest's logic LP be applied to the
practical business of evaluating, or at least coping with, charges of the fal­
lacy of equivocation? Lewis's answer is this: ambiguity is, or could be, per­
vasive in natural language, therefore we can never disambiguate our lan-
268 INFORMAL FALLACIES

guage fully. Hence, according to Lewis (1982, p.439), the best we can do is
to weaken our logic so that it tolerates ambiguity.
Classical logic preserves truth in a valid argument even in the face of
ambiguity in just the following sense — if each ambiguous term in the pre­
misses and conclusion is disambiguated in the same way in every instance,
then the conclusion is true on that disambiguation if the premisses are. But
in practice, that is not much help. Why not? Well in practice, the problem
of equivocation may be posed by the fact that one confronted with the
problem of evaluating an argument may not be able to fully disambiguate
all its terms. Lewis calls such an arguer a "pessimist," meaning I suppose,
that he may be able to disambiguate the terms of the argument bundle he
faces, to some extent, but he cannot perhaps be sure that he has fully dis­
ambiguated every term in the argument bundle. How could such a critic of
argument still reason in the face of the potential ambiguities inherent in
natural language? Lewis (1982, p.440) offers this answer.
If things are as bad as he [the pessimist] fears, he must perforce reason
from premises accepted merely as true-osd, or at best as true-osd only ...
His highest hope for his conclusion is that it will be true-osd only, or at
least true-osd.
The fact that classical logic preserves truth if each term is consistently dis­
ambiguated the same way throughout the whole argument is, therefore, no
help to the pessimist. For he cannot safely make such an assumption of con­
sistency when reasoning in natural language. So he should move to a
weaker sentential logic when confronted by the possibility of equivocation.
To see what this proposal amounts to in practice, let's take the case of
the pessimist who forgoes classical propositional logic for the weaker sen­
tential logic of LP. 5 According to Lewis, the implications that preserve
truth-osd are those of the first-degree fragment of LP. Technically, this
means that in a valid argument, you can never go from premisses that are
true-osd only or mixed to a conclusion that is false-osd only.
But what does this criterion of validity amount to in practical terms? It
seems to amount to this requirement: in a valid argument, if the premisses
are true on some disambiguation, then the conclusion will be true on some
disambiguation. More precisely, it means this: if the premisses are true-osd
only, or true-osd and false-osd, then the conclusion is never false-osd only.
In other words, it says that if the premisses are true on some disambigua­
tions — meaning at least one or possibly all disambiguations, including the
possibility that they are false on some disambiguations as well — then the
 EQUIVOCATION 269

conclusion must be true on some disambiguations (or equivalently: the con­

clusion can't be false on all disambiguations). To sum up: if the premisses
are true-osd, then the conclusion must be true-osd. This would seem to con­
form to what Lewis writes by saying that the implications preserved by LP
are precisely those that preserve truth-osd.
But how exactly does this criterion of validity assist us in guarding
against possible fallacies of equivocation? It is still not entirely clear how it
does provide such assistance. Some examples can help. Let's consider the
case of contraposition.
As Lewis (1982, p.433) points out, contraposition fails for implication
in LP. Where  is both true and false and A is true only, A implies Β ν IB,
but 1(B ν IB) fails to imply 1A. However, contraposition holds for Priest's
conditional, A → B, defined as ΙΑ ν Β. We have then, an interesting ques­
tion for LP as an applied logic. If a conditional 'If A then B' occurs in a sen­
tence, how do we translate it — as 'A → B' or as 'A implies B.' In general,
it would be more appropriate to use A → Β, since that is the general condi­
tional, unless the context makes it clear that logical implication is meant by
the 'If ... then.' So that is the policy we adopt. When we come to the next
relevance logic to be examined however, we will have to adopt a different
policy. For it will have no symbol for 'If ... then' other than implication.
The inference form A → Β |= IB → 1A is valid in LP. Consider an
instance where the term 'bank' could mean 'savings bank' or 'riverbank'
and the term 'plant' could mean 'factory' or 'bit of vegetation.' What can
we say then about the following inference?
(13) If Bob went to the bank then Bill went to the plant.
Therefore, if Bill did not go to the plant then Bob did not go to
the bank.
Suppose the premiss is true on some disambiguation. Suppose, for exam­
ple, that it is true that if Bob went to the savings bank then Bill went to the
factory. Does it follow that the conclusion must be true on some disambigu­
ation? Clearly it does not follow that if Bill did not go to the bit of vegeta­
tion, Bob did not go to the riverbank. But it does follow that the conclusion
must be true on at least one disambiguation, namely this one: if Bill did not
go to the factory then Bob did not go to the savings bank. Hence the infer­
ence is valid in just the sense required by LP. If the premiss is true-osd,
then so is the conclusion. Our problem is this however. How could the LP-
validity of the inference above stop a would-be equivocator from using it to
270 INFORMAL FALLACIES

fallaciously get someone to accept its conclusion?

Perhaps it would help us to look as well at an instance of an inference
that has a form that is not valid in LP (disjunctive syllogism).
(14) Either Bob went to the bank or Bill went to the plant.
Bob did not go to the bank.
Therefore, Bill went to the plant.
Here truth-osd is not preserved. If the first premiss is true only on one dis­
ambiguation of 'bank' and the second premiss true only on the other, the
conclusion could be false on both disambiguations of 'plant'. This inference
is not valid in LP.
Now why is it that the first inference guards against the equivocator in
a way the second one fails to?
Lewis (1982, p.440) suggests that relevance logic has a use in dealing
with equivocation when he writes, "... it may serve a purpose to have a
partly relevant logic capable of stopping fallacies of equivocation even
when equivocation is present." But precisely what is this purpose or use of
the relevant logics? How does it stop fallacies of equivocation? Lewis does
not attempt to answer these pragmatic questions. Nor does he really
attempt to tell us precisely what the "fallacy" of equivocation consists in as
a move in argument. Hence what he offers gives no clear guidance on how
the requirement of relevance "stops" an equivocally fallacious move in
argumentation. Lewis offers a formal mechanism and suggests its possible
application to the fallacy in a helpful and constructive way. But he stops
short of the pragmatic job of carrying the application through as a project
of practical or applied logic to the natural discourse of the usage of
equivocal argumentation in realistic refutations and criticisms.
To make progress in addressing this pragmatic question, we need to
inquire: why is the first inference equivocation-proof in some way the sec­
ond one fails to be? In turn, to answer this question, the best thing is to
look at how LP might function in being applied to the equivocator's move
of presenting a bundle of arguments.
Could adopting the requirement of validity in LP stop an equivocator?
Let us go back to (13), an example of an inference that is valid in LP. The
possible disambiguations of (13) run as follows.
(15) If Bob went to the savings bank then Bill went to the factory.
Therefore, if Bill did not go to the factory, Bob did not go to the
savings bank.
 EQUIVOCATION 271

(16) If Bob went to the riverbank then Bill went to the vegetation.
Therefore, if Bill did not go to the vegetation, Bob did not go to
the riverbank.
(17) If Bob went to the savings bank then Bill went to the vegetation.
Therefore, if Bill did not go to the vegetation, Bob did not go to
the savings bank.
(18) If Bob went to the riverbank then Bill went to the factory.
Therefore, if Bill did not go to the factory, Bob did not go to the
riverbank.
(19) If Bob went to the savings bank then Bill went to the factory.
Therefore, if Bill did not go to the vegetation, Bob did not go to
the riverbank.
(20) If Bob went to the savings bank then Bill went to the factory.
Therefore, if Bill did not go to the vegetation, Bob did not go to
the savings bank.
(21) If Bob went to the savings bank then Bill went to the factory.
Therefore, if Bill did not go to the factory, Bob did not go to the
riverbank.
(22) If Bob went to the savings bank then Bill went to the factory.
Therefore, if Bill did not go to the vegetation, Bob did not go to
the riverbank.
(23) If Bob went to the riverbank then Bill went to the vegetation.
Therefore, if Bill did not go to the factory, Bob did not go to the
savings bank.
And so forth, for all possible subsequent disambiguations. (15)-(18) are the
consistent disambiguations, all valid in classical logic. No remaining disam­
biguations, including (20)-(23) above, are valid in classical logic.
In light of our analysis of the fallacy of equivocation, how could an
equivocator use (13) as a fallacy? To see how, let's suppose that there is a
connection between Bob's going to the savings bank and Bill's going to the
factory, so that in fact the following conditional is conceded as true: if Bob
went to the savings bank then Bill went to the factory. Let's say that once
Bob goes to the bank, then Bill always does the payroll. Let's also suppose
that there are no other relevant connections, e.g. there is no connection
between Bob's going to the savings bank and Bill's going to the vegetation.
272 INFORMAL FALLACIES

Hence all the other disambiguated conditionals are false. In this situation,
how could the would-be equivocator equivocate?
The equivocator would choose an argument that would prove the con­
clusion he needed to prove. Suppose that in the game of dialogue in ques­
tion, the opponent needs to prove the following conditional: if Bill did not
go to the vegetation, Bob did not go to the riverbank. How to "prove" it?
Well, he may be aware that (19) has that conclusion and (19) has a true pre­
miss. Here then is a strategy for the equivocator. Put forward the argument
(13), full, as it is, with ambiguities. Then hope that the respondent will
match up your conclusion to be proven with the premiss he has conceded as
true, in effect selecting the invalid argument (19). But also hope that the
respondent will perceive the form of argument not as that of (19), but as
one of the valid forms (15)-(18), viewing the argument through the valid-
looking bundle (13).
Such an interchange of moves between the respondent and the oppo­
nent conforms to the pattern of the classical fallacy of equivocation. The
respondent is invited to choose both validity and true premisses by rolling
up two arguments into what appears to be one.
Now we come to the main question. (13) is valid in LP, meaning that if
the premiss is true on some disambiguation, so must the conclusion be true
on some disambiguation. Nonetheless, as we saw above, this sort of validity
is no obstacle to the equivocator. True, if the premiss is true-osd, so is the
conclusion. But all the equivocator needs is an argument where the premiss
is true-osd and false-osd, and the false disambiguation implies the conclu­
sion he needs to prove. The fact that truth-osd is guaranteed by the LP-
valid equivocal argument is no bar to the strategy of the equivocator.
The problem is that LP allows you to go from premisses that are true
on some disambiguation to a conclusion that is false on some disambigua­
tion. Hence LP allows an argument like (19) to be included in a bundle of
sentences declared LP-valid. Therefore, LP is no bar to the equivocator.
It seems then that LP is not a particularly useful propositional logic for
games of dialogue where equivocation needs to be stopped. Whether there
are other uses for LP in regard to equivocation remains an open question.
But whatever uses it might have seem limited. If we stick tο LP as our con­
ception of validity, we can never go from a set of premisses that is true on
some disambiguation to a conclusion that is false on every disambiguation.
But that seems a small consolation to the one who has lost the game
through swallowing a fallacy of equivocation. If he accepts a set of premis-
 EQUIVOCATION 273

ses true on some disambiguation as a basis for accepting a conclusion that is

also true on some disambiguation, that is no consolation if the conclusion
also turns out to be false on the disambiguation that counts. Such accep­
tance may lose him the argument even if the implication that was the basis
of it is valid in LP. Hence LP, at least by itself, does not seem to be much
help in aiding an arguer to contend with the fallacy of equivocation.

8. R-Mingle as a Logic for Equivocators

The second suggestion advanced by Lewis (1982) is that the relevant
logic RM (R-Mingle) could be applied to equivocation. According to Lewis
(p.439), the implications that preserve both truth-osd and truth-osd only
are those given by the first-degree fragment of RM. According to Anderson
and Belnap (1975, p.341), the following are the fourteen axioms for RM. 1,
^, and ν are defined as in LP.
Rl A→A
R2 (A → B) → [(B → C) → (A → C)]
R3 A → [(A → B) → B]
R4 [A → (A → B)] → (A → B)
R5 ( ^  ) → 
R6 (^)→
R7 [(A →) ^ (A → )] → [A → ( ^ )]
R8  → (Α ν )
R9  → (Α ν )
R10 [( → ) ^ ( → )] → [(Α ν ) → ]
R11 [ ^ ( ν )] → [( ^ ) ν ]
R12 ( →1) → ( →1)
R13 ll → 
R14  → ( →  )
The propositional logic RM consists of the above fourteen axioms, closed
under the following two rules of inference.
→E: from A →  and A to infer 
&I: from A and  to infer  ^ 
In system RM, the three forms of argument below — respectively ex falso
quodlibet, disjunctive syllogism, and arguing from any premiss to a tautol­
ogy — are not valid.
274 INFORMAL FALLACIES

 ^ ΙΑ Α ν  
 ΙΑ  ν IB

Using the same tables for 1, ^, and ν as in LP we can check for invalidity.
We can see, for example, why the third form of argument is invalid in RM.
Suppose that A is true on all disambiguations and  is true on some disam­
biguations and false on some disambiguations. Then even though A, the
premiss, is true on all disambiguations, the conclusion, Β ν IB, need not be
true on all disambiguations. For if  is mixed then IB is also mixed. Hence
Β ν IB is also mixed. This inference fails to preserve truth-osd only. The
first one, ex falso quodlibet, fails for a different reason. If A is mixed, then
1A is mixed, and hence  ^ IA is mixed. But  could be false on all disam­
biguations. Hence ex falso quodlibet fails to preserve truth-osd. In RM, all
valid inferences must preserve both truth-osd and truth-osd only.
The concept of implication for RM is the same as that of Kalman impli­
cation, as outlined by Makinson (1973, p.32ff.). Conjunction, disjunction
and negation can be defined using the three element model below, adapted
from Makinson (p.38).

The curved arrows define negation. If a sentence A has the value M, then
1A also has the value M. If A has one of the remaining two values, then 1A
has the other one. For example, if A has the value T, then 1A has the value
F. That sums up the table for negation, which is the same as the one we
gave for negation in LP. The order of the model serves to define conjunc­
tion and disjunction. The value of  ^  is always the lower value. For
example if A = Τ and  = M, then  ^  = M. Or to take another exam­
ple, if A = M and  = M, then  ^  = M. The value of Α ν Β is always
the higher value. For example if A = Τ and  = F, then Α ν  = T. Or,
to take another example, if A = F and  = M, then Α ν  = M. Again,
the tables for  and v , so defined, are the same ones we gave earlier for
LP. 6 The model above is therefore a convenient way of summing up the
logical constants of LP and RM.
 EQUIVOCATION 275

The same model can be used to express the idea of implication for RM.
As we remarked above, implication in RM preserves truth and truth-only.
That is, we can never go validly from premisses that are true-only to a con­
clusion that is false-only or mixed. And we can never go from premisses
that are true-only or mixed to a conclusion that is false-only. We could sum
up this concept of implication as below.

Never goes Preserves truth-only.

Never goes Preserves truth.

Of course, in the context of equivocation, 'truth-only' means 'true on all

disambiguations,' 'false-only' means 'false on all disambiguations,' 'mixed'
means 'true on some disambiguations and false on some disambiguations,'
and 'truth' means 'true on at least one disambiguation.'
Although RM is a relevant logic, it can be shown that the following
argument form is valid in RM.

We can see why by considering the following table.

A ֊  | 1A | IB |   IA Į B v ĮB
~ M F | M F  M
M T M F ' M T
M M M M M M
T T F F F T
 ғ ғ  ғ 
F   F F T
F F   F 
M F M  M 
F M |  | M I F I M
Looking under the column for   IA, we see that it never gives   value.
Hence truth-only will be preserved, in the sense that we can never go from
  1A with the value  to B v IB with the value M or F. Also, truth is
276 INFORMAL FALLACIES

preserved. For we have no row where  ^ 1A has the value M and Β ν IB

has the value F.
Now we come to the → that occurs in all the axioms of RM. Clearly A
→  cannot be defined as ΙΑ ν Β, the way the conditional, →, was defined
in LP. For by that definition, modus ponens would not be valid. Where A
is mixed and  is false-only, ΙΑ ν Β and A both have the value M, but 
has the value F. It seems therefore that the → in the Anderson and Belnap
axioms for RM should be taken as a metalogical symbol for implication, as
we have characterized it above.
Here then is a key difference between LP and RM. When we say, for
example, that modus ponens fails in LP, we mean that the conditional pre­
miss of the form A →  is a conditional, and → is a logical constant that can
be defined in a table, like any other logical constant. When we say that
modus ponens is valid in RM, we mean that the premiss A →  expresses
an implication relationship between A and B.
However, we do not wish to pursue the logical development of RM any
further here. Mainly we want to see whether it could be any use, as Lewis
suggests, in helping us to deal with equivocation as a fallacy. For this pur­
pose it is enough, at least for the present, to read → as 'implies,' meaning
that a valid implication preserves truth and truth-only. We should
remember that we are working with the first-degree fragments of LP and
RM. This means we are excluding from consideration any formula where
one arrow occurs within the scope of another. Therefore, transforming an
implication concept into a conditional defined by a table that can be
generalized to all occurrences of 'if ... then' in the language may be no tri­
vial job. As Makinson (1973, p.40) notes, one of the best ways to carry out
this job is to introduce a necessity operator first, and then define the condi­
tional in the modal logic that results. For the present, we will not take this
step. Instead, let us turn to the question of the usefulness of RM for study­
ing equivocation.

9. RM and Equivocation
Consider the following variant of the standard example of equivoca­
tion.
(24) If death is the end of life then death is the perfection of life.
Death is the end of life.
Therefore, death is the perfection of life.
 EQUIVOCATION 277

The form of this inference {modus ponens) is not valid in LP, but it is valid
in RM. 7 Under what dialectical circumstances could (24) be a fallacy of
equivocation?
Suppose White advances (24) in the course of an argument where he
needs to get Black to accept the conclusion of (24). Black, let's say, accepts
the first premiss as true under its more plausible disambiguation: if the
death is the goal of life then death is the perfection of life. Suppose Black
also accepts the second premiss under its more plausible disambiguation:
death is the termination of life. But suppose that Black does not find either
remaining disambiguation of either premiss plausible. Suppose then, given
these assumptions about Black's position, White tries to use (24) in the
equivocal way as a bundle of arguments to try to get Black to concede the
conclusion of (24), an implausible proposition for Black. What White has to
do if he is to succeed in pulling off the equivocation is to roll up the valid
disambiguations of (24) — all of which have implausible premisses —
together with the invalid disambiguation above that has premisses both
plausible for Black. This strategy is one that is based on a fallacy, of course.
Where both premisses are disambiguated, as above, to come out plausible
for Black, the conclusion does not follow by valid argument in classical
logic. Hence the equivocator must push for (24) in its non-disambiguated
form as a collection of sentences that can be disambiguated in various ways.
And (24) in that form, is not really an argument at all, in the sense of being
a univocally designated set of propositions. Still, White hopes that (24)
looks to Black to be an argument, a valid single argument. His hope is pin­
ned on the surface appearance that (24) looks to have a form that is valid in
classical logic {modus ponens).
But now let us consider modus ponens as a form of argument in RM.
What does it mean to say that A implies  (A → B) in RM? It means two
things. First, it means that if A is true on all disambiguations, then so is .
Second, it means that if A is true on some disambiguations, then so is . So
let's assume that A is true on all disambiguations. Then by the first require­
ment, so is  true on all disambiguations. Let's assume that A is true on
some disambiguations. Then by the second requirement, so is  true on
some disambiguations. Hence modus ponens must be valid as a form of
inference in RM, meaning that it preserves truth-osd and truth-osd only, in
every instance.
So let's now consider a game of dialogue where the logical rules are
given by RM rather than by classical logic. How does the equivocator fare?
278 INFORMAL FALLACIES

At first impression, things look good for the equivocator now. The
form of (24) is valid in RM. So it seems he can advance (24) with impunity.
But let's see how White could deploy (24) against Black in an RM-
based game of dialogue. First, since (24) is valid, both Black and White
have to concede that if the premisses are true on some disambiguations,
then so is the conclusion. All White has to do then is to get Black to con­
cede that all the premisses of (24) are true on at least one disambiguation.
But let's look at them. Certainly the second premiss is true (plausible to
Black) on some disambiguation. That is, Black will accept the proposition
that death is the end (termination) of life. However, the first premiss is not
true (plausible to Black). For the antecedent of the first premiss is true on
some disambiguation and false on others, but the consequent (Death is the
perfection of life) is unambiguously false, as far as Black is concerned.
Hence the first premiss is unacceptable for Black, according to the require­
ments for implication given by RM.
But second, since (24) is valid in RM, both Black and White have to
concede that if the premisses are true on all disambiguations, then so is the
conclusion. But the problem here for White is that the second premiss,
'Death is the end of life' is not true on all disambiguations. It is false if 'end'
is disambiguated as 'goal,' so far as Black is concerned.
If RM is the model of validity, (24) is valid. But if RM is the standard
of validity, then Black has to reject one premiss or the other as implausible.
RM has two requirements guaranteed by a valid argument — it preserves
truth-osd and truth-osd only. But using one requirement, one premiss is
implausible, and using the other requirement, the remaining premiss is
implausible. Hence Black could not accept (24) as a reasonable argument
for the conclusion 'Death is the perfection of life' if he cleaves "logically" to
the standards of RM and to his prior presumptions of what is plausible.
This example suggests then that RM might be useful as a defensive
screen for the arguer who is confronted with an equivocal argument in
dialogue.
But does Black do any better with RM than where the model of impli­
cation is that of classical logic? In classical logic, the bundle (24) is valid on
some disambiguations, but always has an implausible premiss on any valid
disambiguation. In classical logic, Black is safe from equivocation if he
demands both validity of argument and plausibility of all premisses. So
Black is relatively secure, already, with classical logic as a defence against
equivocation. Why is he any better off with RM? I think the answer here
 EQUIVOCATION 279

has already been suggested by the approach of David Lewis, and it can be
brought out in the present case as follows.
When confronted by (24) in a game where classical logic is the model
of valid argument, Black has to disambiguate. He has to know that 'end'
has a definite, known set of disambiguations, and he has to check all of
them out. If there are no disambiguations that yield both plausible premis­
ses and a valid argument, he can securely reject the equivocal argument as
a fallacy of equivocation. But in realistic arguments, the potential for dis­
ambiguations may be limited by the known lexicon of White or Black. This
open-endedness of disambiguations is especially pertinent where vague
terms may be part of an argument in natural language. In that case, an
arguer can never be entirely secure against a strategy of equivocation.
However, where RM is the model of implication Black does not have
to disambiguate. In this context, the argument is no longer composed of a
set of propositions that are true or false, as in classical logic. In classical
logic, Black is confronted by a bundle of arguments, where the "bundle" in
reality may be open-ended, and perhaps not even defined exactly in his
given lexicon. Equivocation, then, can be a very real danger.
But where the model of valid argument is RM, the argument is made
up of a set of given sentences, each of which is true on all disambiguations,
false on all disambiguations, or true on some disambiguations and false on
others. Here Black is given a definite fixed set of sentences as the "argu­
ment," and his evaluations range over the disambiguations of these sen­
tences in a way that can make them complete and final, even if the set of
disambiguations has never been precisely and completely designated by
the arguers. Here Black does not need to disambiguate. He can apply the
test for validity directly to the sentences themselves, even if they are
ambiguous.
Some care is needed in applying RM, however. From the given infor­
mation that the premisses are separately true on some disambiguations, it
need not necessarily follow that the conclusion is true on some disambigua­
tion, according to the standards of RM. That inference could be a fallacy.
What we can correctly infer is that if the premisses are true on some disam­
biguation — that is, jointly true on at least one particular disambiguation,
the same one throughout all the premisses — then the conclusion must be
true on some disambiguation (not necessarily the same one). However, as
(24) illustrates, the premisses could be true on some disambiguations or
other — differing from one premiss to another — but that tells us nothing
280 INFORMAL FALLACIES

at all about whether the conclusion is true on some or all disambiguations.

To so infer would be a sort of fallacy which could perhaps, with reason, be
called a form of equivocation. But if we are careful in applying RM rightly,
then this sort of fallacy can be excluded.
So it may be well to repeat the warning. In an RM-based game of
dialogue, from the given information that some of the premisses are true on
some disambiguations and some of the premisses are false on some disam­
biguations, the respondent may not necessarily infer the conclusion that the
premisses are jointly true on some disambiguations and jointly false on
some disambiguations.
Another warning is that validity is not enough, for the purposes of
dialectic, to secure an arguer of protection against equivocation in every
respect. If he thinks of accepting the conclusion that a sentence is true on
all disambiguations, he needs to check — even if the argument is valid in
RM — whether the premisses are true on all disambiguations in relation to
his position in the argument. If he thinks of accepting the conclusion as true
on some disambiguations, he needs to check out the premisses in that
regard as well.
In short, RM requires of a respondent that, if this conception of valid­
ity is to be used in practice, he should check to see whether some or all dis­
ambiguations of the premisses have been considered before he moves to
accept the conclusion of a valid argument. If the respondent is careful
enough to always demand this requirement be met, he can never be a victim
of a fallacy of equivocation within RM in just these senses. Having met the
requirements of validity and truth-osd only of the premisses, he can never
become committed to a conclusion that is false on any disambiguation.
Having met the requirements of validity and truth-osd of the premisses, he
can never become committed to a conclusion that is false-osd only. To that
extent, RM is equivocation — proof.
Just as classical logic can be used to stop equivocation in one way, RM
can be used to stop equivocation in a somewhat different way. Where the
rules of the game are those of classical logic, the defender against the
equivocal argument can reply: "You want me to take 'end' in one premiss
of (24) one way, and in other premisses another way. You can't have it both
ways. Be consistent in your disambiguations or I won't accept your argu­
ment." If the attacker is consistent, at least one premiss must be false, even
if the argument is classically valid. If one premiss is false (implausible), the
defender need not accept the premisses. Ergo, equivocation as a fallacy can
 EQUIVOCATION 281

be defended against in classical logic. But to make the defence work, the
logic has to apply to the disambiguated propositions.
Where the rules of the game are the relevance logic RM, the defender
against the equivocal argument can reply: "You want me to take 'end' in
one premiss of (24) one way, and in the other premiss another way. On
those different disambiguations both premisses are true. Moreover, the
argument (24) is a valid form of argument according to the rules of the
game. But before I accept the conclusion as true-osd, you are going to have
to prove to my satisfaction that the premisses are jointly true on some dis­
ambiguations. If one premiss is true on one disambiguation, and the other
is true on another disambiguation, that may not be good enough. You must
cite further disambiguations to make your point. You must show me that
the premisses are all true on some disambiguations. Or alternatively, if you
want me to accept that the conclusion is true on all disambiguations, you
must show me that the premisses are also true, as a joint set, on all disam­
biguations." This rejoinder can stop equivocation in the case of (24). In this
respect then, RM does offer an effective way of stopping equivocation.
We shouldn't overlook the other aspect of RM as a screen against
equivocation, namely that it rules invalid some forms of argument that are
valid in classical logic. Consider the following argument.
(25) Death is the end of life or life goes on forever.
Death is not the end of life.
Therefore, life goes on forever.
Let's say that White is set to prove the conclusion of (25) above. Black
takes the opposite side of the argument, against immortality. However,
let's say that Black accepts the first premiss as plausible as disambiguated
one way (death = termination) but not the other. And he accepts the sec­
ond premiss as plausible, disambiguated the other way (death = goal), but
not the first way (death = termination). How does RM determine his
responses to White's advancement of (25) as an argument?
First, note that in RM, if A is mixed and  is false-only, then Α ν Β
has to be mixed (disjunction always has the greater value in the three-ele­
ment model). Therefore, if disjunctive syllogism is valid, then Black has to
accept the conclusion of (25), namely the sentence that life goes on forever,
as being true on some disambiguations. Why? Well, first of all, Black
accepts that death is the termination of life. Hence he must accept that
'Death is the end of life' is true on some disambiguations. Hence he must
282 INFORMAL FALLACIES

accept the first premiss of (25), because if one disjunct is true on some dis­
ambiguations, then the whole disjunction must be true on some disambigu­
ations. But Black must also accept that the second premiss of (25) is true on
some disambiguations. For if a sentence is true on some disambiguations,
then so is its negation true on some disambiguations. Now by the concep­
tion of validity in RM, if you accept a set of premisses is jointly true on
some disambiguations, as these premisses are accepted by Black, then you
have to accept that the conclusion is true on some disambiguations. There­
fore Black, if he plays by the logical rules of equivocation games, has to
accept that life goes on forever, if disjunctive syllogism is valid in RM.
Luckily for Black however, disjunctive syllogism is not valid in RM, as
we can see by the following table.
A  1A A ν Β
Τ M F Τ
M τ M Τ
M M M Μ
τ τ F τ Table for
τ F F τ Disjunctive Syllogism
F Τ Τ τ
F F Τ F
M F M Μ Χ
F M Τ Μ
In the second to last row, both premisses Α ν  and ΙΑ, have the value M.
But the conclusion, B, has the value F. And of course, this combination of
values violates the requirement of RM that in a valid argument, truth on
some disambiguations is always preserved.
Hence it is that White could use (25) as an equivocal bundle of argu­
ments in classical logic, but he cannot use (25) as a valid argument at all in
RM. In certain cases then, RM functions as a screen against the
equivocator by rejecting the argument altogether. In RM, Black would be
secure against any attempt to use an argument like (25) as an equivocation,
merely by playing consistently according to the standards of validity
imposed by RM. (25) would not get through the screen at all.
Ultimately, however, RM permits the same kinds of instances of
equivocation as LP. For it remains true that in a valid argument in RM, you
can still go from premisses that are true on some disambiguations to a con­
clusion that is false on some disambiguations. Essentially the same sort of
 EQUIVOCATION 283

equivocal argument as indicated by (19) can also be mounted in RM. How­

ever, RM does have one advantage. You can screen against equivocation
entirely by tightening up your standards for the premisses, by requiring that
the premisses be true on all disambiguations. On that requirement, the con­
clusion can never be false, in an RM-valid argument, on any disambigua­
tion. At least in that respect, RM is somewhat more useful than LP as a
screen against equivocation.
Both LP and RM have the advantage of telling you that if the premis­
ses are true on some disambiguation, then so is the conclusion. But that
assurance is of limited helpfulness in guarding against the equivocator. It
does tell you that if the premisses are true on some disambiguation then the
argument may not be entirely worthless, in that the conclusion must be true
on at least some disambiguation. But it is no proof against equivocation, for
it may still be that the conclusion is also false on some other disambigua­
tion. However, at least in RM you can be guaranteed that the conclusion is
never false on any disambiguation if the premisses are true on every disam­
biguation.
Although RM seems the better system for defending against the
equivocator, it has some technical problems. Going back to (24), we might
have our doubts about translating the conditional as 'implication.' More
work is needed here to make RM better applicable to conditional reason­
ing.

10. Conclusions
Equivocation can arise because a sentence can express more than one
proposition. Basically, an argument is a set of propositions. But in argu­
ment in natural language, one arguer can present a set of sentences where,
because of ambiguity, the person to whom the argument is directed may
select out various different sets of propositions (arguments) from the set of
sentences advanced by his opponent. Therefore, in dialogue in natural lan­
guage, equivocation is a constant danger because ambiguity, and its com­
panion vagueness, may always be present in the sentences of the language.
This much of the dialectical framework of argument makes equivocation
theoretically possible in dialogue, even where the rules of classical logic are
observed so that the arguers agree on the rules that define when a conclu­
sion follows validly from a set of premisses.
From a point of view of applied logic, equivocation is more than just a
theoretical possibility when a context is present in argument that makes
284 INFORMAL FALLACIES

some sentences more plausible when disambiguated one way and other sen­
tences more plausible when disambiguated another way. The recipient of
the argument, whose position makes one disambiguation more plausible
than another, may then be tugged two different ways in relation to a set of
sentences. This dual tug may produce a realistic fallacy of equivocation
where the recipient is pulled one way — towards plausible premisses — and
at the same time another way — towards reconstructing an argument that
comes out valid. This binary pull is what makes a strategy of equivocation
realistically effective in games of dialogue.
The most interesting realistic cases of equivocation occur where
abstract words like 'compatible' or 'obligatory' occur — as in the cases cited
by Cederblom and Paulsen (1972) and Hamblin (1970) — because such
words are susceptible to shades of vagueness, or multiple disambiguations
of subtle kinds, that can easily ensnare the most careful and attentive
arguer into problems of interpretation. In order to adequately understand
problems relating to equivocation in these cases, it is necessary to think of
argument as two-person positional dialogue. To begin to understand
equivocation as a fallacy, we need to ask how an arguer should appeal to
procedural rules of dialogue when confronted by a bundle of arguments
advanced by his opponent, where this bundle can or should be criticized as
equivocal. The problem then is one of finding rules or procedures to screen
out equivocal argumentation.
There are two basic approaches to this problem. The first one is to
study equivocation from the outside. This involves reconstructing the
dialectical context of an equivocal move in argumentation, and finding
dialectical rules or procedures that police equivocation, or at least give an
arguer some grounds for criticizing equivocation as a response to an
equivocal move. The other strategy is to study equivocation from the inside
by modifying the conception of 'valid argument' built into the game.
Instead of having a classical logic of propositions to define 'valid argument,'
we move to the level of sentences and adopt a relevance logic. The receiver
of argument, having made this shift, no longer needs to break down the
vague or ambiguous bundle of sentences into well-defined sets of proposi­
tions in a separate step of evaluation. He simply applies the appropriate
relevance logic directly to the possibly ambiguous set of sentences advanced
by his opponent. In our treatment of equivocation, we have combined these
approaches, trying to take each of them as far as it can go in helping us to
work towards a solution to the basic problem of equivocation.
 EQUIVOCATION 285

The first relevance logic we examined, LP, by itself cannot be an effec­

tive defensive screen against all equivocations. The problem we found was
that LP allows you to go from premisses that are jointly true on some dis­
ambiguations to a conclusion that is false on some disambiguations. That
still leaves enough room for an equivocator to conduct an effective strategy
of equivocation, in some cases. Although it blocks equivocation somewhat
better than classical logic in some cases, LP cannot block every equivoca­
tion, and seems to leave significant possibilities for equivocal arguments
open.
Then we examined RM. The logic RM also lets you go from premisses
that are true on some disambiguations to a conclusion that is false on some
disambiguations. And contraposition is valid in RM for the → that repre­
sents implication, the only conditional we have in RM so far as we devel­
oped that system of sentential logic.
One potentially useful way to define the conditional in RM could be
carried out in terms of Makinson's three-element model as follows. A con­
ditional always comes out false on all disambiguations if it goes from a
higher element to a lower element in the model. The conditional comes out
true on all disambiguations if the antecedent is false on all disambiguations
or if the consequent is true on all disambiguations. These two clauses define
all nine rows of the table, excepting only the row where both antecedent
and consequent are mixed. There, we say that the conditional is mixed.
A  A→B
Τ M F
M Τ Τ
M M M
τ τ Τ
τ F F
F Τ Τ
F F Τ
M F F
F M Τ
This conditional is reflexive, non-symmetrical, and transitive. Among the
forms of inference that come out valid for it are MP, conjunction, and con­
traposition.
With this conditional, a better case can be made out for cases like (24)
as ways of enabling an arguer to defend against equivocation. However, it
286 INFORMAL FALLACIES

remains to be seen whether it is formally feasible to define the conditional

above as part of RM. I take it from the efforts of relevance logicians to
avoid defining a conditional for RM by a three-valued table that there are
formal problems in the undertaking. So it remains to be seen whether my
suggestion will turn out to be useful.
I conclude that the approach to the outer analysis of equivocation as a
fallacy is the more fundamental and, so far, the more promising. The inner
approach has the very attractive advantage of removing the need to disam­
biguate sentences into propositions, but still does not entirely exclude
equivocation as a fallacy. Moreover, the inner approach appears to have
some technical problems needing further development. At present then,
the understanding of both approaches seems to offer the best hope of gain­
ing further insight into the fallacy of equivocation. The basic fact about
equivocation is that you can never be sure that you are excluding it as a pos­
sibility, as long as you are using a natural language with empirical terms
that are not defined tightly for all possible contexts. You can exclude it if
your sentence logic is RM. But to use RM in an argument, you must accept
that your premisses are true on all disambiguations. And once again, in a
natural language, it may be only in the exceptional type of case that you
could have such an assurance.
In natural language, it may be that in the broad spectrum of cases we
are most likely to encounter in the practice of argumentation, there can be
no really effective assurance that the possibility of equivocation is excluded.
But perhaps for practical purposes, this level of assurance can be waived.
Even if the potential for equivocation may be omnipresent, we seem to
get in trouble with it only where plausibility tugs us to interpret one sen­
tence in an argument one way, and another sentence a different way. It is
the factor of this dual contextual tug that leads us to the characteristic prob­
lem of equivocation — the urge to fall into the confusion of accepting many
arguments as if they were one argument. If so, perhaps the most practical
way to guard against equivocation is to check each term that occurs in more
than one sentence in an argument, and see if a shift can be detected. To
make the argument be valid yet have plausible premisses, are we required
to interpret some term or phrase in two different ways at its two separate
occurrences in the argument? This seems the most practical test for equivo­
cation.
This test for equivocation turns on the question of how each partici­
pant in the argument most plausibly disambiguates the sentences in the
 EQUIVOCATION 287

argument. This is a question of the arguer's position in the context of

dialogue. Therefore, in the most practical terms, it may be that the outer
approach of the pragmatics of dialogue is the most basic and important
aspect of the management of equivocation in natural language argumenta­
tion.

NOTES

1. One could possibly even take the extreme view that a man and a woman are never compat­
ible. According to Engel (1976, p.64), the following quotation may be attributed to G.K. Ches­
terton: "If Americans can be divorced for "incompatibility", I cannot conceive why they are not
all divorced. I have known many happy marriages, but never a compatible one. For a man and
a woman, as such, are incompatible".
2. 'Mixed' here means true on some disambiguations and false on some disambiguations.
3. See also the comments on nested conditionals in Makinson (1973, p.40).
4. In a subsequent article, Priest (1980) is led to question this definition of the conditional,
and finds it too limited.
5. Priest gives a quantifier logic for LP as well, but we confine ourselves to the case of propos-
itional logic here.
6. We could define v(A ν ) = max(A,B) and v(A ^) = min(A,B).
7. (24) is valid, at least, if the conditional of the first premiss is translated as the → of RM.
 CHAPTER 11: INFORMAL LOGIC AS A DISCIPLINE

There are numerous recent indications in North America of a strong

emerging need for a pedagogy of applied or informal logic as a coherent dis­
cipline. In the field of education, numerous articles expound the need for
'critical thinking' as a potentially useful skill to bettter equip students at all
levels with reasoning skills that are often deficient. At post-secondary
levels, many new courses are being taught on informal logic or critical
thinking. A California state law even requires the teaching of such a type of
course at all state universities.1 The "new wave" textbooks, characterized
by an applied and topical study of realistic argumentation, and less formal
logic than has been traditional, continue to appear in even greater num­
bers. 2
One can easily appreciate the inherent usefulness of a well-developed
subject of informal or applied logic study in any democratic society with
voting, elections, a court system with legal advocacy, a free media, and an
open policy of social decision-making. Optimal social policy in a free soci­
ety surely requires a rational appraisal of arguments for and against a posi­
tion. Moreover, rational dialogue is surely the only available means of con­
ducting open argumentation and negotation, preferable alternatives to
military force as the means of settling disputes. As a method of conflict
resolution, informal logic ultimately represents the only way of conducting
reasonable dialogue, perhaps ultimately the only workable alternative to
repression, propaganda, or even war, as means of resolving disputes or
enforcing a point of view.
Some of the goals of critical thinking include critical empathy, the abil­
ity to understand another's — perhaps opposed — point of view on a posi­
tion. Another goal is the ability to look behind one's own position to see
what lines of argument it could be based on, rather than fixing on a dogma­
tic or uncritical tenure of one's favorite view. A third goal is the ability to
make finer discriminations in the statement of a position so that conditions,
exceptions and refinements in it can be made as an intelligent response to
argumentation. Anyone who has had the experience of marking student
290 INFORMAL FALLACIES

essays, on virtually any topic, will recognize that none of these goals is
pedagogically trivial.
All of these goals and aspirations are noble and worthwhile. They do
characterize the promise of philosophy as a productive discipline. No doubt
the study of philosophy, as it is currently practiced, is one of the most effec­
tive ways to pursue these goals. But how far are we towards developing a
clear and coherent methodology of practical logic that we aspire to have as
the vehicle for meeting or supporting the goals? Although the textbooks
designed to teach the skills needed for a discipline of practical logic have
both multiplied and improved in recent years, there are reasons to question
whether they are based on a well-grounded knowledge of their subject-mat­
ter. There remain many fundamental controversies about how we ought to
proceed in this area.
The foregoing chapters have conveyed some idea of the state of the art
of practical logic, but even the fact that our first chapter was entitled "A
New Model" indicates that the field is in a rapid state of change. In fact, my
own feeling is that while many of the textbooks are going in good direc­
tions, their efforts are not well-coordinated. And indeed, the thrust of this
book is to consolidate the foundations of the field to some extent, in order
to provide a resource for future textbook authors, among other readers.
My own resume of the current state of this field — especially in regard
to the fallacies — has now been made. And the theory of argument devel­
oped therefrom has laid out the lines of advance. But since the field is frag­
mented at present, it may be important to spend another chapter examining
some of the main controversies about informal logic currently subject to
dispute.
The basic problem, in my view, is that arguments as they occur in "real
life" are extremely complex. Any particular argument needs to be
approached as a potentially complex phenomenon that can outrun an
attempt to impose theoretical order on it, in a number of directions. Practi­
cal logic, as an applied undertaking, must never lose sight of the fact that as
an applied discipline, it confronts objects of analysis that are inherently
vague, unruly, and difficult to pin down.
That problem is not insurmountable, for practical logic any more than
any other applied discipline. But it does require a certain care in approach­
ing the subject. We have seen that in very many instances, "fallacies" turn
out to be milder forms of criticism than outright refutation. Indeed, they
are often criticisms that can be supported, or in other cases defended
 INFORMAL LOGIC AS A DISCIPLINE 291

against. This approach suggests that practical logic often does not go so far
as to tell us that the argument is wholly good or wholly bad. It should have
the more limited objective — more soberly construed — of throwing some
light on the weak points or the stronger points in a particular argument.
However, it is well to observe that the views stated above are controv­
ersial. To see how, let us review some of the recent arguments.
A current controversy is whether the individual subject areas like
economics, law, paleontology, and so forth, ought each to have its own crit­
ical thinking unit specifically for that topic, or whether there should be a
separate critical thinking course for all disciplines. Another big issue is the
extent of formal logic in such a discipline. Some practitioners devote little
or no space to formal topics like propositional calculus.3 However most
texts concede that mathematical structures of varying sorts — not by any
means always classical logic exclusively — should be a supportive element
of theory integrated with the study of actual argumentation. What sort of
formal logic is also an issue? Obviously, the view favored by the present
author is that many nonclassical logics and models of inference are enorm­
ously useful. Yet another question is the extent to which such formal mod­
els are idealizations that need to be supplemented by pragmatic steps to be
filled in before we reach the level of realistic disputation. Yet another ques­
tion is the role of structured models of argument other than propositional
logics, for example, games of dialogue.

1. The Role of Formal Logic

Some see formal logic as inapplicable to serious conversational
argumentation. This view often perceives formal logic as trivial or sterile,
an artificial, mathematical construction that is only very obliquely related,
if at all, to reasoning in natural language. 4 The disanalogies between the
classical material conditional and the natural language 'if ... then' are often
cited as evidence for this argument.
Many formal logicians have themselves encouraged this perspective.
For example, Tarski has suggested that natural languages are inherently
contradictory and vague because they lack the proper resources to even for­
mulate their own truth-conditions for the statements made in them.
According to Tarski, only a formal language like mathematical logic can
avoid the contradictions involved in its own semantic closure. 5 Many for­
mal logicians are inclined to think that coherent conditions of 'correct argu­
ment' can only be formulated in an abstract, formal system. How these for-
292 INFORMAL FALLACIES

mal conventions "translate" into natural language, or specific instances of

argumentation, is a very chancy business, perhaps the job of linguists,
social scientists, or other empirical scientists.
On the other hand, many of those who expound informal logic as a dis­
cipline reject the applicability of formal logic to actual argumentation not
because of the ineffability of argumentation, but because of the inability of
formal logic to adequately model realistic arguments. By these lights, we
should forget about formal logic as so much "busy work," and get on with
the real job of analysis of real arguments. 6
Both of these extreme positions have severe problems. The problem
with the "informal" approach is that once all formal procedures are cast
aside, the issue of whether an alleged fallacy is incorrect or not becomes
itself open to argumentative disputation. With no normative guidelines or
objective models, we are cast back to the level of the subjective quarrel and
adversial debate. Whoever can argue the loudest or strongest carries the
day. Hence there is no useful thing to say about whether the original argu­
ment is good or bad.
The problem with the "purely formal" approach is — as we have seen
— that many of the fallacies defy formalization within classical logic. Thus
at least on a narrow or conservative view of what constitutes formal logic,
we are bereft of resources applicable to adequate models of correctness/
incorrectness for the fallacies. Thus neither the purely formal nor the
purely informal approaches are useful or productive.
We think a more salutary approach is exemplified by the treatment of
our previous chapters, where our criteria of "formal" are widened to
include nonclassical formal models like many-valued logic, relatedness
logic, and graphs of arguments. By allowing these different models of argu­
ment a place in the theory of dialogue, one's resources for the study of fal­
lacies is encouragingly widened. It becomes clear that at least partially for­
malized models of argument can be made applicable to the major tradi­
tional fallacies in an enlightening way. And the theory of games of dialogue
enables us to draw together these formal resources in a general framework.
Of course much of the traditional doom and gloom about the fallacies
is justified if we do analyze them at the ad hoc level of free quarrels or
debates. Such adversarial, unregulated, subjective models can never take
us very far towards offering objective guidelines to fairly determine
whether a given argument is correct or fallacious. But the advent of the
dialectical model of argument can overcome the intrinsic shortcomings of
 INFORMAL LOGIC AS A DISCIPLINE 293

the laissez-faire, survival-of-the-fittest approach to argumentation. Hence

the value of pure informality versus that of rigorous and pure formalism is,
at bottom, a non-issue. A subtler and more serious evaluation of the state
of the art should have made it clear by now that the question is how formal
or strict the rules of reasonable dialogue should be.
The real problem of the formality or informality of applied logic should
be recast as the question of the relative strictness of the rules of dialogue.
What is the right balance of freedom and regulation in question-asking and
answering in games of dialogue? That is the important question.
We have found that the requirements of a theory of dialogue that can
most fairly model realistic interchanges in the context of the fallacies
include (a) strictly regulated procedures that (b) leave the participants free-
dom and latitude in making their moves. Of course, (a) and (b) are incon-
sistent with each other. But that sort of tension between (a) and (b) is typi-
cal of regulated games of strategy. To have an interesting game, you must
combine (a) and (b) in the right mixture.
The big theories of dialogue we have appealed to show a scale of rela-
tive strictness (tightness) in how they regulate games of dialogue. The
Lorenzen games are regulated by a very tight degree of strictness. The
atomic (simple) propositions have their truth-values determined by "exter-
nal facts" and then the complex propositions can be algorithmically deter-
mined in their truth-values by strictly finite procedures. The polar opposite
of the strict necessity of dialogue is the political debate, where virtually any-
thing goes. The Hamblin and Hintikka games lie somewhere between these
two poles.

LORENZEN GAMES
HINTIKKA GAMES
HAMBLIN GAMES
DEBATE (QUARREL)

The Hamblin games are relatively free and unstructured. A player can
always answer 'No commitment' to a question with relative freedom. The
objective of "information exchange" in the game is not defined in a way
that tightly defines the player's win-strategies. The Hintikka games are
more strictly regulated. They are cumulative — a player cannot retract
commitments — and a player does not have a 'No commitment' option in
294 INFORMAL FALLACIES

answering questions. Moreover, a win-strategy for each player is precisely

defined — the player wins who first proves his thesis from the other's com­
mitments. Now the general problem for the analysis of the fallacies is this.
What is the right balance of freedom and regulation in question-asking and
answering?
The Walton games fit in between the Hamblin and Hintikka games.
Like the Hintikka games, the objective of the player is to first prove his
thesis from the other player's commitments. Hence win-strategy is precisely
defined. However, like the Hamblin games, the players generally have a
'No commitment' option in answering questions. But the use of this option,
in a Walton game, is circumscribed by the previous commitments — in par­
ticular, the dark-side commitments — of the answerer. So the Walton
games are more strict than the Hamblin games, but more open than the
Hintikka games.
The issue is not whether to embrace formal logic wholeheartedly or
reject it altogether. The issue is to find the right level of strictness or free­
dom in pragmatic structures to best model the process of rationally con­
ducting and evaluating the give-and-take of realistic argumentative inter­
change in dialogue. We need to find the right level of theory to fairly rule
on criticisms and replies in a reasonably realistic type of situation. The
games I have advocated are designed to work towards this level in relation
to the informal fallacies.
Future research needs to study the right question-answer mechanisms
in games of dialogue for fair and reasonable management of arguments. I
hope now, at least, we can see better where to go from here.

2. Dialectic as a Theory of Argument

We have seen, following Hamblin, Hintikka, Rescher, and Lorenzen
that games of dialogue can be studied as purely formal as well as realistic
processes. Thus formal dialectic offers the most promising vehicle for the
future study of argumentation and fallacies.
It is very useful here to make a distinction — now widely familiar in the
linguistics of discourse processes — between semantics and pragmatics.
Semantics has to do with truth and falsity, logical implication and consis­
tency, and related matters. Pragmatics has to do with conversations and
other interpersonal processes that find their medium in the continuity of
discourses on issues or topics of dialogue. Surely without contradicting
ourselves we can make sense of the idea that pragmatics as a study may be
 INFORMAL LOGIC AS A DISCIPLINE 295

in some sense "less formal" than semantics. If so, the notion of an "infor­
mal logic" can make sense.
But is "informal logic" a misnomer? Would it not be much less mis­
leading to speak of "pragmatic dialogue theory" or perhaps "theory of
argument" as the appropriate vehicle for the study of fallacies, criticisms,
and other concepts of argument? In linguistics, the term "argumentation" is
now often used to indicate the pragmatic nature of the discipline of studying
realistic samples of discourse. Unfortunately, the term "informal logic"
suggests that the pragmatic study of argument is casual, or has nothing
whatever to do with formal logic. Both these views are badly wrong, and
harmful for the study of argument as a serious discipline.
Dialectic, like any scientific model of the real world, involves a certain
degree of obstruction if it is to be expressed as a theory. Such abstraction
need not be in itself a bad thing if it leads to consistency and organization
in a theory. The theory of argument should be abstract, but that does not
mean that it cannot be or should not be applicable to realistic argumenta­
tion on issues of importance. Fortunately for the theory of argument, real-
life arguments do benefit from clear procedural conventions of dialogue
articulated by the participants as rules of permissible moves in argument.
An organized theory of these sets of permissible rules in games of dialogue
is therefore applicable to realistic argument in an edifying way.
Dialectic does all the many things that we practical students of argu­
ment want, like showing us that there is always another side to the argu­
ment on any important issue. But dialectic does these things best if it
already includes logic, just as pragmatics is built around semantics.
However, if formal logic is to be included within dialectic, the question
remains: which formal logic? Classical logic has a central place in logic, yet
nowadays there are many non-classical logics that are equally formal, and
also have some claim to being called "formal logic." The multiplicity of
logics today makes one wonder whether formal logic is itself the tidy discip­
line that its informal logic detractors often seem to reject for fear of its
rigidity. But we have found it appropriate to use non-classical logics to
model dialectical contexts for the various fallacies.
Another objection concerns the pluralism of different models of 'cor­
rect argument' embodied in analyses of different fallacies. We utilized clas­
sical PC, non-classical propositional logics like relatedness logics and many-
valued logics, inductive concepts of argument, and aggregate theory,
among other models. We also observed that a problem can arise where an
296 INFORMAL FALLACIES

argument turns out to be valid by analysis of one of these logics, yet invalid
by another. Can such a pluralism be contended with by a single, coherent
discipline?
The brunt of the answer to this question has to be borne by dialectic.
Formal dialectic can incorporate different conditionals, different concepts
of proof, within its over-arching structure as a conception of argument. The
answer then is that disputants must agree on the game they are playing, the
rules of procedure, and the notion of conditional that defines the appropri­
ate concept of correct argument. From the dialectical point of view, an
argument is correct insofar as it meets the standards of correctness stipu­
lated by the game. And these standards may vary given the particular con­
text of the question at issue, and the nature of the dispute. Thus the flexibil­
ity of formal dialectic to allow different formal games of dialogue allows for
the applicability of dialectic to realistic disputations in actual argumentation
in natural language.
However, this very flexibility could pose worries of pluralism in games
of dialogue we have adopted in studying the fallacies. We basically started
with Hamblin's framework and then introduced the idea of win-strategies,
thereby moving towards Hintikka's framework of logical dialogue. These
games of dialogue have clearly formulated procedural rules, and hence we
were able to analyze the fallacies as violations of these rules. Nevertheless,
a certain conventionalism was implicit in the project, for the participants in
a game of dialogue could agree to different kinds of rules in different con­
texts. The only necessity is that once rules are adopted, they should be
enforced where they apply.
Even so, further study needs to be done on the appropriateness of dif­
ferent kinds of rules for the different fallacies. Moreover, there remain cer­
tain key differences in the kinds of games we proposed as models of the fal­
lacies. ABV was a relatively simple game that had only one kind of ques­
tion permitted, namely the yes-no question. CBV, on the other hand, per­
mitted why-questions as well. Why this difference? Could we not standar­
dize games of dialectic, or at least explain why some are more complex than
others?
The dark-side commitment-store was a key feature of all the dialogue-
games we constructed as models of the various fallacies. Another key fea­
ture was the win-objective that each player must prove his own thesis from
his opponent's commitment-set. These two features were optional, and
could be varied in some of the contexts we discussed. But on the whole,
 INFORMAL LOGIC AS A DISCIPLINE 297

they were relatively basic and central to the study of the fallacies. However,
the kinds of questions allowed in the games was a matter we found better to
leave relatively open. It is now time to see why.
Basically, the reason is that there is some uncertainty about the func­
tion of why-questions in games of dialogue. Why-questions were introduced
by Hamblin, and for some time I followed the practice, in my research on
the fallacies, of incorporating them into games of dialogue. However, I
came to realize that following this practice made it much more difficult to
define 'presupposition' and 'direct answer' for questions in the games.
Also, problems relating to the petitio principii fallacy as studied in Walton
(1984) evidently traced to some uncertainties concerning the function of
why-questions. Hence it is time to re-think the function and role of why-
questions in logical games of dialogue as a general problem.
Should why-questions be retained, and what do they really mean?
These are key questions for future studies in the theory of argument.

3. Function of Why-Questions
The why-question in a game of dialogue is essentially an imperative
that requires, in relation to its answerer: "Prove proposition A! That is,
give me some proposition  that can function as a basis of proof for A."
This type of question is called a "challenge" by Mackenzie (1981) meaning,
I suppose, that the answerer is challenged to supply some proof or justifica­
tion for his stance. The Hamblin games include why-questions as well as
yes-no questions. Allowing generally for both kinds of questions is under­
standable, given the information-orientation of the Hamblin games. How­
ever, once the win-loss rules of the games are tightened up, one begins to
question more sharply the function of each type of question in the game of
dialogue.
There are various positions one might reasonably take concerning the
alleged purpose of why-questions in games of dialogue. The various pos­
sibilities are given below. The first two proposals are named for two of my
students who suggested these functions in a class discussion.
SHINFIELD PROPOSAL: The function of the why-question is to get a
commitment (or a set of commitments as premisses) of the questioner that
implies (imply) the proposition queried.
This proposal represents a view of the why-question that would make it
express the question "Why should I be led to accept A?" as the equivalent
298 INFORMAL FALLACIES

of "Why A?" In other words, "Prove A" means "Give me some proposition
 such that A is a consequence of  and  is a proposition I am committed
to!" In some contexts of dialogue, this construal of the why-question seems
natural. But in other contexts it seems artificial or inappropriate.
In some contexts of argumentation, a why-question 'Why A?' could be
more naturally interpreted as meaning 'Why do you accept A?' Hence a
second interpretation.
STRANG PROPOSAL: The function of the why-question is to get a com­
mitment (or a set of commitments as premisses) of the answerer that
implies (imply) the proposition queried.
This proposal represents a view of the why-question as a request to the
answerer to provide a justification for the answerer's acceptance of A, the
proposition queried.
Here then we have two sharply contrasting interpretations, each of
which seems plausible in its own right.

It seems then that 'Prove A!' can have two different meanings: (1) 'Prove A
to me!' (Show that it follows from something I accept), or (2) 'Prove A
from your own position!' (commitments you accept). The second meaning
comes out as: show that A follows from something you accept.
A third possibility exists. The why-question could combine the two
proposals above in a single request. This approach gives us the Hamblin-
rule (Rl) of Hamblin (1970, p.271). We refer to the rule Hamblin calls "un­
necessarily strong."
HAMBLIN PROPOSAL: The function of the why-question is to get a
commitment (or a set of commitments or premisses) common to both
answerer and questioner that implies (imply) the proposition queried.
Which of these proposals is appropriate depends on the purpose of the
game of dialogue one has in mind. If the purpose of the game is the friendly
objective of exchanging information, or informing each other on the basis
 INFORMAL LOGIC AS A DISCIPLINE 299

of some core of common agreements, then the Hamblin proposal could be

appropriate. The premissary base needed for this purpose is the common
commitment-stock of both parties.
However, if the purpose is persuasion of the other party, then the
objective of the game is to prove your own thesis from premisses accepted
by the other party. Whether you yourself as questioner accept these premis­
ses is beside the point. In that context, the Strang proposal is clearly the
most appropriate account of the function of a why-question.
If the objective, or a main objective of the game is to defend one's
position against attacks, challenges or refutations of the opposing party,
then the Shinfield proposal comes to the fore as the best model of the func­
tion of the why-question. Here "Prove it!" means "Defend it against my
objections or skeptical queries by showing that it is a part of your position!"
Here the answerer is proving in the sense of defending his own position.
While all three of the above possibilities can be shown to make good
sense if fleshed out, there are reasons to believe that each of them is some­
what superficial in relation to the realistic functions of dialogue, at least as
stated above. Hamblin himself recognizes that his own rule (Rl) may be
too strong, because in realistic dialogue, a demand for a one-step proof is
too harsh. It may be enough to satisfy the demands of a reasonable why-
question if the answerer eventually comes up with the right sort of commit­
ment after a longer series of questions and answers. If so, a satisfactory
answer to a why-question may be, in many instances, more of a promissory
note than an outright delivering of the ultimate premiss that furnishes the
proof the proposition queried. This suggestion invites various weaker for­
mulations of the function of a why-question in dialogue.
The weakest possible rule is that the answerer is required to produce
any pair of propositions, ,  → A in response to the question 'Why A?,'
never mind whether  or  → A are commitments of either party or not.
The philosophy behind this proposed rule is to give the answerer enough
freedom to pursue a lengthy chain of proof. This proposal makes sense if
the win-loss rule of the game demands that the answerer, if he is to win,
must go back up a chain of modus ponens inferences until he eventually hits
on a set of premisses that are commitments of the questioner. When he
does get to these "ultimate premisses" his strategy of winning is assured.
Perhaps the roles could be reversed and the questioner would win if he
asked questions leading through a series of valid steps of inference in the
answers culminating in premisses that are commitments of the answerer. As
300 INFORMAL FALLACIES

with the Strang proposal and the Shinfield proposal, the roles could be con­
sidered in reverse in different contexts of dialogue. But the distinguishing
feature of this new approach is that the rule for answering is much weaker.
It does not require that the propositions immediately furnished as an
answer to the why-question be commitments of either party. Instead, the
win-loss rules of the game dictate working towards such commitments as
the objective of the game. To win the game, the questioner — or perhaps
the answerer — should lead towards answers that are commitments of the
party he needs to obtain to win the game.
In designing the dialogue-rule for answering why-questions, there are
four variants we could have as options, in accord with the proposals above.
In answer to a why-question, the answerer should be able to opt out various
ways by replying 'No commitment' in some form. But if he opts to respond
by offering premisses that ostensibly serve to prove the proposition
queried, then there are four possible forms this part of his response could
take. That is, if a questioner asks 'Why A?' and the relevant rule of the
game for providing proofs is modus ponens, the answerer must produce a
pair of propositions ,  →A at the next move such that (i) there are no
restrictions on  and  → A at all, or (ii) B and  → A must be commit­
ments of the questioner, or (iii)  and  —» A must be commitments of the
answerer, or (iv)  and  → A must be commitments of both parties.
The first option (i) may seem too weak. In dialogue,, the participants
often may not be clear about where a series of questions is leading. But
there could be fallacies of irrelevance committed if there is no control over
the direction of a series of questions at all. Hence one might want to con­
sider variants weaker than (ii), (iii) or (iv), but stronger than (i). For
example, one might require that  should be related to A, even if  and 
→ A may not be commitments of either party. This way of restricting the
relevance of answers to why-questions requires only that A and  have
some subject-matters in common. By this way of regulating answers, the
answerer must not stray off the topic. He must keep his propositions
offered in proof at least related to the proposition queried in the question
last asked.
Another possibility of strengthening (i) in an interesting way would be
to require that the information in  be outside the information in A. That
is,  has to tell you "something new" in addition to the information already
given in A. For both ideas of relatedness and informational inclusion, mod­
els of valid argument other than classical propositional calculus could be
 INFORMAL LOGIC AS A DISCIPLINE 301

proposed. We have in mind the relatedness logics and dependence logics of

Epstein (1981) in these contexts.
Hamblin included the yes-no questions and the why-questions both in
the same game of dialogue. But as one begins to formulate the objectives of
the Hamblin type of game more sharply the function of the why-question in
the game becomes a little hard to fathom. It seems that separating out the
why-questions and the yes-no questions in different games could be quite
important. Intuitively, one can appreciate the idea that why-questions or
requests for justification are important in dialogue where the object is to
understand the cut and thrust of defence and refutation that is clearly part
of the problem of many of the major informal fallacies. But the precise
import of the why-questions seems open to many different interpretations,
each with some reasonable claim to legitimacy.
To some extent then, the precise rules of each game remain open to
interpretations of the objective of dialogue in a particular context. This
means that the job of applying a model of logical dialogue to a particular,
realistic case of argumentation has to remain a matter of judgement to
some extent. This gap between abstract models and particular cases of
realistic disputation is not, in itself, always a bad thing, however. For as we
have seen repeatedly, many real disputes take place in a context where
there is no prospect of rationally evaluating the arguments unless many fac­
tors, e.g. the topic, the conclusions of the disputants, and so forth, are
defined or clarified.
As we also saw, with applicability to practical disputes comes also a
certain gap. The convèrsational quarrel and rhetorical debate are to a
goodly extent unmanageable by purely formal or normative rules of proce­
dures because, in the reality of the argument marketplace, their aim is to
defeat the adversary, if necessary, at the expense of logic, or even of argu­
ment altogether. Thus the actual examples we have analysed, from par­
liamentary and other debates and instances of argumentation, needed con­
siderable "cleaning up" before dialectical methods of analysis could be
effectively brought to bear. This gap between the actual disputation and the
normative regulation of how the argument should be, if it is to be correct,
is always present. It is a fundamental oversight to ignore it.
Thus it is a basic insight that in the gap between formal models of argu­
ment and actual, raw argumentation there is an intermediate pragmatic
level of filling in. The nice thing about dialectic as a theory however is that
it allows these gaps to be reasonably filled in, and by the right means. If we
302 INFORMAL FALLACIES

come to a chunk of raw argumentation with premisses, deductive links,

rules and so forth missing or unstated, how can we fill in the blanks? Too
often the Standard Treatment of fallacies has allowed such filling in to be
done in an unsystematic way that may in fact be unfair to the original
advancer of the argument criticized (usually not around to defend himself).
However, in dialectic, the fundamental question must always be: "What
would the arguer say on the basis of his position in the game of dialogue or
in the sequence of questions and answers in a continuation of the game of
dialectic?" Anything other than this sort of dialectic evidence for commit­
ments of the arguer must be presumption of dark-side commitments. Prag­
matics comes in to study the giving of justification for such presumptions.
Thus the use of actual examples as specimens for analysis, while not in itself
the final word on analysis of fallacies, can be highly instructive in teaching
us how to fill in these gaps. Thus a pragmatic orientation is a welcome
direction for the future development of informal logic as a discipline.

4. Subject-Specific Nature of Arguments

One interesting objection to the possibility of informal logic as a discip­
line concerns the topic-sensitivity of argumentation to a specific field. This
argument works from the premiss that a principle of argument may be cor­
rect in one field but incorrect in another. For example, according to
McPeck (1981, p.72), a principle of reasoning in business or law might be
fallacious in science or ethics. According to McPeck, the notion of informal
logic is therefore impaled on a dilemma. Learning to reason in isolation
from specific subject areas takes us no further than formal logic, where
forms of argument abstract from the specific content of the variables. But
learning to reason substantively involves learning about the actual subject
areas of the content, and therefore requires the breadth of learning of a
Renaissance man. According to McPeck (1981, p.81), the informal logician
is inconsistent in wanting to have it both ways.
Some of the sting is taken out of this objection by the elementary
observation that not only are there different principles of reasoning within
different fields, but there are even different principles of reasoning within
the same field. True, a valid principle in law might be fallacious in science.
But even more to the point, a valid principle in criminal law may be invalid
in civil law. A valid principle in applied mathematics may be invalid in some
branch of pure mathematics. Rather then, we would conclude that varia-
 INFORMAL LOGIC AS A DISCIPLINE 303

tions in principles of correct argument occur within different dialectical con­

texts, whether in the same field or others.
Even with some of its sting removed however, the objection raises
some interesting questions. If a certain amount of pragmatic filling-in is
required in the adjudication of an allegation of fallacious argument in a dis­
cipline, it may be that this filling-in can only be done by an expert within
that discipline.
For example, suppose a medical doctor is accused of committing an
equivocation on the word 'death.' The critic alleges that in one premiss it
means 'irreversible cessation of cardiopulmonary function' and in another
premiss it means 'irreversible cessation of all brain function.' Suppose
further that the physician replies, "I didn't equivocate. Irreversible cessa­
tion of cardiopulmonary function is one criterion of irreversible cessation of
all brain function. That's the way we doctors define death, namely as 'ir­
reversible cessation of all brain function.' The cessation of cardiopulmonary
function is just one indicator of death, for if the whole brain ceases func­
tioning, the cardiopulmonary function inevitably ceases in a short period
anyway. Thus I didn't equivocate — I used a different criterion for death,
but did not use the term 'death' in two distinct meanings." How can the cri­
tic — who is, let's suppose not a medical doctor — dispute this rebuttal?
The doctor's special knowledge in the field — let's say he is a neurologist
who specializes in the determination of death — allows him to decisively
reject the allegation of ambiguity.
Of course the critic could reply that the definition of 'death' is not a
purely medical matter, but rather a question of ethics. Although this reply
shifts the topic from one field to another, perhaps more accessible to the
critic, the physician may still reply that he used the term in the medical
sense, or that his definition is defensible within ethics as well, and so forth.
This sort of problem is not just newly remarked upon. Sextus
Empiricus, in Outlines of Pyrrhonism (II, §236) makes this very objection
(also quoted in Hamblin, 1970, p.95).
As regards all the sophisms which dialectic seems peculiarly able to
expose, their exposure is useless; whereas in all cases where the exposure
is useful, it is not the dialectician who will expose them but the experts in
each particular art who grasp the connection of the facts.
This dilemma has already been rebutted in chapter seven, where we saw
that, in many cases, the expert's sayso should and can be challenged
304 INFORMAL FALLACIES

through reasonable procedures of questioning in dialectic. The necessity of

such dialectical reasoning has been amply demonstrated in the trial
practice of cross-examination, where many a confident expert's testimony
has rightly had its credibility undermined by skilled dialectical examination.
However, the problem posed by Sextus has, at its roots, several
genuine theoretical questions for dialectic. What parts of the resolution of
a questionable move in disputation are up to the critic or evaluator, and
what parts are up to the disputants themselves to fill in, possibly from the
backgrounds of their own expertise in a field, or from other resources?
Future studies of argumentation in informal logic should keep this problem
in mind in designing different games of dialogue.
Mitroff and Mason (1981) think of dialectic as a process with three par­
ticipants — two opposed disputants and a referee who evaluates the
dialogue. However, the model of dialogue due to Hintikka that we have
been adopting has no use for a referee. In a Hintikka game of dialogue, the
rules determine the outcome of the game with no need for a third-party
evaluator.
But the problem is that as critics or informal logicians who are to rule
on fallacies, we do often find ourselves in the role of a third party who is
supposed to be "neutral" in taking a position in a given dispute. So the
problem is — when does the critic himself become a party to the dispute?
Or if he doesn't, what sort of role does he play in the game of dialogue?
We had already seen the same sort of problem begin to emerge in
chapter 5 where, in the second type of enthymeme, the critic is in the situa­
tion of filling in missing premisses where the arguer is not present to accept
or reject the proposed premisses. The "critic" had to look at the
enthymeme both from the point of view of its proponent and also from the
point of view of the audience it was designed to convince. The problem is
that the critic is neither fish nor fowl. How can his function be understood
in games of dialogue?
My best suggestion is that the critic is the one equipped to understand
the theory of argument and the rules of reasonable dialogue. But how does
he apply them to the particular case of a real argument, without taking sides
and himself becoming a participant in the argument? I think he can do this,
but his evaluation needs to be premissed on the game of dialogue having
been fruitful enough to reveal the participant's objectives and commitments
in relation to a particular dispute.
 INFORMAL LOGIC AS A DISCIPLINE 305

5. Case Studies on Circular Reasoning

It does sometimes happen that substantive disputes take place between
experts in a particular field of scientific inquiry where the dispute turns on
some methodological point or issue pertaining to practical logic. Where this
situation develops, it may well transpire that a non-expert third party —
perhaps a philosopher, or at any rate a party somewhat removed from the
special field in which the dispute arose — may very helpfully serve as an
external commentator or analyst of some critical general contours or points
of contention in the dispute. Such a third party, perhaps partly by virtue of
his neutrality or non-involvement — can adopt a broader perspective, and
thereby sometimes succeed in throwing some light on key moves of argu­
ment crucial in obtaining a better understanding of the real nature of the
dispute or controversy. In effect, this third party may play the role of an
"argument analyst" or practitioner of what we have been calling practical or
informal logic.
The most likely reason that this sort of helpful relationship has not
been too evident in the past is that informal logic lacks a well-established
methodology, or even a general theory of argument and criticism studied as
an academic subject in its own right. I hope, of course, that this monograph
will help to support an effort in this very direction.
Even despite the infancy of the needed methodology, some case
studies can be cited where a third-party non-expert perspective would be,
or has been helpful in yielding some insight into certain methodological
problems in research in special fields. Some excellent examples relate to
controversies about certain arguments in fields of scientific research that
have been criticized as viciously circular. The problem is whether these alle­
gations of fallacious argument can be sustained, or whether the circles are
of a non-vicious sort.
Our first case study of an allegation of petitio principii pertains to the
geologist's practice of using rocks to date fossils and also using fossils to
date rocks. The branch of paleontology called stratigraphy plots the layers
of fossil remains embedded in vertical rock strata and arrives at conclusions
about the relative ages of these fossils on the basis of their spatial findings.
Some laypersons, and even some geologists, have suspected a vicious circle
in this process of reasoning. For example, Rastall (1956, p.168) states this
criticism as follows.
306 INFORMAL FALLACIES

It cannot be denied that from a strictly philosophical standpoint geologists

are here (in using fossils to determine relative ages of strata) arguing in a
circle. The succession of organisms has been determined by a study of their
remains buried in the rocks, and the relative ages of the rocks are deter-
mined by the ages of the organisms they contain.
The allegation stated here voices the criticism that the circle inherent in the
stratigrapher's process is a vicious one that destroys the reliability of his
conclusions.
Responding to these charges, Harper (1980) has argued that if we pay
close attention to the way paleontologists actually do use the evidence of
spatial findings to arrive at temporal conclusions, there is no vicious circu-
larity. Harper's reconstruction of the actual process of inference used by
the stratigrapher breaks that process down into two steps. First, the super-
position of strata is used, at the individual local section, to infer the relative
age of fossils. This first step, according to Harper (1980, p.246) is a relative
and local judgement: "... fossils are not dated apart from the strata that
contain them; when we infer a relative age for a particular local fossil or
fossil assemblage, we simultaneously infer the same age for the local strata
which contain it, and vice versa."
The second step takes place when the stratigraphie paleontologist
looks for orderly succession of fossils over a whole region. That is, his
investigation here may be horizontal as well as vertical, one supposes
Harper to be suggesting. Then at this second stage of investigation, if regu-
lar non-random patterns are found, the paleontologist uses these patterns
to infer relative ages for both the fossils and also the strata that contain them.
It is helpful to construct a graph of the process of the paleontologist's
argument as reconstructed by each of the disputants in this controversy. First,
Rastall's statement suggests the following reconstruction of the process.

Order of Relative ages

succession of of the rocks
fossils buried containing
in rocks. fossils.

Why is this circular process vicious or fallacious? It is not clear, but perhaps
the charge of circularity is by itself suspicious enough to throw the burden
of proof onto the stratigraphie paleontologist to defend his methodology.
Harper's defence reconstructs the process in a different way, by distin-
guishing between two findings, namely local succession of fossils and fossil
 INFORMAL LOGIC AS A DISCIPLINE 307

assemblages over a region. His view of the argument could be graphically

represented as follows.

Superposition of Superpositon of
fossils over a strata over a
local section. local section.
 

Relative ages Relative ages of

of strata over fossils over a
a local section. local section.
Regular non-random
patterns over a
region of fossils.

Certainly the above graph of Harper's reconstruction of the sequence of the

paleontologist's argument shows no circle.
Harper's defence is the fourth step that a defender can take in response
to a criticism of petitio principii, if you remember the seven steps in our
analysis of petitio criticisms and rebuttals given in chapter six, section
seven. Harper is reconstructing the paleontologist's argument in a different
way than Rastall has. The premisses and conclusions are different, and
there is no circle at all in Harper's more sophisticated version of the paleon­
tologist's line of reasoning. But Harper, in so rebutting, is also introducing
a defence somewhat similar to our sixth stage of evaluation. For he argues
that both the relative ages of the strata and fossils are evidentially based on
an external premiss providing independent evidence for both of these find­
ings, namely the evidence of non-random patterns over a whole region of
fossils.
Clearly the dispute between the arguments of Rastall and Harper is a
specialized controversy within the science of paleontology. It is for the
stratigraphic paleontologists to settle. Does the practical logician have any
light to throw on it? Does he have any place butting in at all? It is my con­
tention that the informal theorist of argument not only can, but has actually
been, of some help here.
308 INFORMAL FALLACIES

Gratifyingly for the present author, Professor Harper cited Woods

and Walton's analysis of the petitio (1975), and deployed that analysis as his
vehicle for his defence against the charges of petitio levelled against paleon­
tology in this instance. It is quite clear that the informal logician will not
and should not have the last word in this type of case, but it is encouraging
that his analysis can be some help.
Further evidence of the legitimacy of argument analysis as a field of
research in its own right, apart from the specialized fields of inquiry or con­
troversy that it may be applied to, is furnished by the fact that disputes on
petitio essentially similar in form occur in other specialized fields of science.
This suggests a recurrent pattern in many separate fields of inquiry. This in
turn suggests that a certain distance from the particular subject-matter by a
third-party referee or evaluator of the arguments may have a genuine and
useful place. Another case study, this time from the field of evolutionary
biology, may help to bear out my claim.
The origin of this case study is a charge of vicious circularity levelled by
Charles Darwin (1859) against his predecessors in the study of taxonomy.
Darwin proposed that his predecessors had argued in a circle when they
claimed that important organs never vary, but decided which organs were
important on the basis that they did not vary. According to Darwin's criti­
cism, the conclusion of his predecessors to be proven is the proposition 'Im­
portant organs never vary.' Presumably then, the evidence needed to prove
this conclusion consists of a set of propositions put into the form of the fol­
lowing inference: organ χ is important, therefore organ χ does not vary.
However, Darwin's criticism continues, the actual evidence marshalled by
his predecessors was a set of propositions of this form: organ χ does not
vary, therefore organ χ is important. In effect then, Darwin accused his pre­
decessors of arguing in the circle below.

Subsequently however, the evolutionists themselves were accused of argu­

ing in a circle. Initially it was not Darwin himself who was so accused, but
other evolutionists. But as the criticism of circularity in evolutionary
 INFORMAL LOGIC AS A DISCIPLINE 309

taxonomy gained momentum, even Darwin's arguments came under suspi­

cion.
These charges of circularity in evolutionary taxonomy have been
reviewed and clarified by David L. Hull (1967), a philosopher of biology.
According to Hull's argument, the charges of circularity, once properly
understood, can be seen to be based on philosophical preconceptions of the
methodology of biology as a science that can be refuted.
The problem arises because evolutionary biologists must start with
some classifications or taxonomies of the organisms they propose to investi­
gate. Using such a taxomony, a biologist can then set to examine the evi­
dence of fossils and then arrive at generalizations about the evolution of the
organisms. However, no system of classification is perfect, and as the evi­
dence of fossils is considered, refinements or alterations in the system of
initial classifications is bound to take place. Indeed, Hull argues that this
process of successive approximation and refinement is properly characteris­
tic of good hypothesis formation and confirmation in scientific method. But
the problem arises out of the fact that some critics find this process circular.
For after all, the taxonomy is being used as an essential part of the interpre­
tation of the fossil evidence. But then the investigation of the fossil evi­
dence is being used to correct and to validate the taxonomy as a useful
structure of classification. Hull breaks this general suspicion of circularity
down into three specific charges.
The first criticism Hull examines (p.177) is the charge that homologous
resemblances (structural resemblances, e.g. where both species have an
arm structure) are defined in terms of phylogeny (common line of descent),
and then phylogeny is defined in terms of homologous resemblances. Hull's
reply to the criticism is (a) to concede that there is a circle in this case, (b)
but to argue that the circle is not vicious. Hull concedes (p.178), "... any
evidence to the effect that a particular resemblance is homologous would
necessarily be evidence to the effect that it was due to common line of des­
cent and vice versa." But he goes on to make the additional point (b) that
the evidence from which phylogeny is inferred can come from another
source as well — any evidence of a dissimilarity between two organisms,
e.g. genetic evidence, could be good evidence for the right way to construe
phylogeny without depending on homology at all. His rebuttal takes this
form.
310 INFORMAL FALLACIES

Clearly then Hull's reply above is a rebuttal that takes the form of the
seventh step. 7 in our procedure for evaluating a circular argument.
The second criticism Hull examines is the charge that, in some cases, a
taxon, or classification of an organism, can begin to assume a permanence
and reality out of proportion to the tentative basis on which it was originally
established (p. 178). Hull's response to this criticism is to concede that there
is a feedback process at work in the successive approximation of use of
taxonomy in evidence, but that the circularity of this process is not falla­
cious. He points out (p.179) that, in real life, certain processes are circular.
For example, as a diabetic gets more overweight there is more insulin in his
blood. But the more insulin he builds up, the more he tends to eat, and con­
sequently he stores up more fat. Similarly, evolutionary taxonomists reclas­
sify in order to present evolutionary descent with greater accuracy. But
according to Hull (p. 179), there is no logical fallacy in either of these circu­
lar processes.
Hull's reply to this charge is to concede that initial classifications in sci­
ence are tentative, and are improved through a "groping" process that
takes place as additional evidence is collected. But according to Hull, all
science starts from imperfect hypotheses which are then refined and
improved through the collection of further evidence. But this process of
successive approximation is not a vicious circle, or evidence of any logical
error or fallacy in the method.
The third criticism is that there are gaps in many fossil records, and
 INFORMAL LOGIC AS A DISCIPLINE 311

therefore some classifications must be made on an insufficient basis of

phylogenetic evidence. Hull's reply here is to point out there is a difference
between two types of criticisms: (1) claiming that an argument is based on
insufficient evidence, and (2) claiming that an argument is viciously circu­
lar. In effect, the distinction Hull alludes to here is the one described by
Woods and Walton (1975) as the difference between bereftness of evidence
and circularity of argument. This is an important distinction, often con­
fused, and Hull is quite right to deploy it here and to insist on its impor­
tance in studying any criticisms of circularity in argument.
Curiously, Hull's analysis of scientific method in the role of taxonomy
in hypothesis formation suggests that there could be a circularity implicit in
inductive methods, parallel to Mill's argument that all deductive reasoning
contains a circle. At any rate, his analysis shows how a third party can
throw light on disputes over alleged fallacies in reasoning amongst experts
in a specialized discipline.
What needs to be done in further studies of these interesting cases is to
reconstruct the whole network of argumentation using the method of
graphs in chapter 6, and then decide whether the model of argument incor­
porates a plausibility requirement that allows or bans the circle. It has
already been shown to some extent that Harper's and Hull's contentions
that the circles they identify are benign, can be justified by the structures of
argument analysis we have set.
Once the specialist's arguments are in, it is up to the dialectician to find
whether the defence of benign circularity can be justified or not. However,
in realistic argumentation, in some cases it may be better to think of the
argument as an open-ended, information-oriented game of Hamblin's type
rather than a strategically closed game of Hintikka's type. If so, it may be
that the dialogue can be continued until the dialectician can get more infor­
mation. Clearly however, as many of our case studies have shown, even a
provisional assessment of an argument can be very revealing and useful.
The usefulness of dialectic as a method of discourse analysis and con­
troversy resolution has been brought out by the uses we have made of it in
the preceding two detailed case studies. A briefer consideration of another
example may also be helpful.
A similar type of dispute concerning circularity has taken place within
the methodology of the social science on the study of the subject of juvenile
delinquency. Kornhauser (1978) distinguishes two types of models of delin­
quency theories, the mixed model and the control model, pictured as
graphs below.
312 INFORMAL FALLACIES

Mixed Model

Lack of Control

Ineffective
Conventional Gang
Institutions Delinquency

Autonomous
Delinquent
Subculture

This mixed model, clearly circular, can be contrasted with the recursive
model below.

Control Model

Organized
Social Adult Crime
Disorgan­
ization

Delinquent
Companions

Social Characteristics Delinquency

of Individuals
 INFORMAL LOGIC AS A DISCIPLINE 313

According to Kornhauser (1978, p.70), nonrecursive models, like the mixed

model, are essentially circular. That does not mean they are fallacious, for
as Kornhauser hastens to add, their circularity may be warranted by the
operation of feedback mechanisms. However, with nonrecursive models,
we may question whether we can talk as favorably of "evidence justifying a
conclusion." Thus Kornhauser concludes that recursive (non-circular) mod­
els are preferable because of their greater power and simplicity.
The case studies from paleontology, evolutionary biology, and sociol­
ogy all show the same general structure in their concern about the coher­
ence of sequences of reasoning in their methodologies relative to certain
problems. We conclude that while the substantive study of the actual issue
in each case is indeed particular to the special knowledge and expertise of
the discipline, still the structure of the petitio argumentation in both is so
clearly the same in its essential form that the study of petitio principii argu­
ment is separate from both subjects. In all instances we had to look at the
graphic network of the over-all sequence of argumentation. In all instances
we had to determine what are the appropriate conditions governing plausi­
ble inference, and thereby reach a judgement of the acceptability or falla­
ciousness of the circles in the argumentation. Each case is a different argu­
ment in a different specialized field, but the same essential job of analysis
needs to be done in order to resolve the question of petitio. These are the
common elements.
But in determining what are the premisses and conclusions of the argu­
ments, and in determining how the various special terms are to be defined,
it is the expert disputants themselves in the particular fields we should best
turn to. Moreover, in seeing how the argument will evolve as a result of the
informal logician's interposition by the pointing out of circular argumenta­
tion, we may again turn to the topical experts themselves. These determina­
tions do not rule out the usefulness of the critic who may, even from a pos­
itional external to the particular discipline of the dispute, step in to clarify
some aspects of the structure of the argumentation. Our thesis has been
that the traditional fallacies most often come out not as "fallacies" at all, in
the sense of knock-down refutations. More intelligently reconstructed in
the theory of logical dialogue, they most often come out as criticisms that
can be replied to, or as requests for procedural clarifications.
We conclude that the study of petitio principii should not be located
exclusively within a particular subject area like sociology, biology, or
paleontology, but is to some extent, properly a distinct object of study for
314 INFORMAL FALLACIES

argument analysis or informal logic in its own right, as a theory of reasona­

ble criticisms.
However, the subject-specificity objection of Sextus is an interesting
one, for it has been made clear that analysis of actual bits of argumentation
does involve pragmatic steps of working with raw argumentation as well as
formal models, dialectical rules, and structured procedures. The specific
skills needed here involve skills of filling in missing steps in a discourse, and
further development of applied logic must surely strive to further study
these pragmatic aspects of argument analysis. Of course, as we have seen,
the dialectical model of argument is the underlying structure. The saying
"There are always two sides to an argument" is always the best guideline in
approaching an argument. In filling in the various pragmatic blanks, we
have seen that the best guidelines are furnished by a careful tracking of the
dialogue sequence and commitment-store information of the participant to
whom the argument is attributed. Then this information can be evaluated
by the rules of dialogue found appropriate.

6. Conversational Pragmatics
Dialectical games are idealized models of how argumentation should
take place, and in many arguments that are analysed retrospectively, the
participants aren't available to answer further questions. Consequently,
application of dialectical models to actual argumentation often requires cer­
tain presumptions of accommodation. By such allowances, conversational
context can sometimes determine whether a statement is true or not.
An example is the slippery slope argument. If someone is on the bor­
derline of being bald, whether or not the statement that he is bald is
adjudged true in a particular context depends on the standards of accuracy
appropriate to that context. However, if the statement is close to being
true, then in most conversations we are inclined to give the speaker the
benefit of any doubt, and simply accept it as true. In numerous conversa­
tional contexts, this tolerance of vagueness leads to no serious problems.
However, as we saw with the slippery slope fallacy, sometimes toler­
ance of vagueness without specifying standards of accuracy does lead to
problems. Over a whole sequence of steps of modus ponens argumentation,
standards of accuracy that were adequate in the beginning become
insufficiently accurate to prevent us from going from truth to falsehood by
ostensibly valid argument. Thus what is usually accurate enough for most
conversational contexts can become insufficiently accurate for another. As
 INFORMAL LOGIC AS A DISCIPLINE 315

David Lewis (1979, p.181) points out, standards of precision are not only
different for different conversations, but can change from one conversation
to another.
As Lewis (1979) relates, this sort of contextual shift sponsors some
skeptical arguments. The skeptic argues that hardly anything is flat. You
think that the pavement is flat, for example. But the skeptic can always get
you to concede that your desk is flatter. If flat is an absolute term, then if
your desk is flat, the pavement can't really be flat after all. Reason: if some­
thing is conceded to be absolutely flat, then nothing else can be flatter than
it.
Lewis's answer to the skeptic's argument is that the latter is changing
the "conversational score." Under the initial standards of accuracy, the
pavement is flat, or close enough to being flat, its unevenness not being
counted as significant relative to the given standards adopted. However,
bringing in comparison with the flatness of the desk raises the standards of
accuracy of the conversation. What was true, or close enough to being true
in the first context became false in the second context. Hence we can see
that truth or falsity is accommodated by conversational conventions that
may vary.
Lewis gives us a picture of how conversational rules work by an anal­
ogy to scorekeeping in a baseball game. In a baseball game, there are rules
which specify how scoring takes place, rules that specify what counts as cor­
rect play, and directives to the players to follow the rules of correct play
and to try to make the score change in certain directions. These rules are
literal or non-accommodating, unlike conversational score-keeping. In con­
versation, according to the rule of accommodation for presuppositions, if
something is said at some point that requires presupposition ρ to become
acceptable, then ρ becomes presupposed after that point, other things being
equal, if it was not already presupposed. This example of a rule of conver­
sational scorekeeping shows how rules in conversations have an accom­
modating character. In baseball, no such leeway is allowed, as Lewis's dis­
cussion shows. If a batter walks after three balls instead of four, the rule of
baseball does not allow for an accommodation, making it true that there
were really "four" balls instead of three. The rules of baseball make for lit­
eral scorekeeping, unlike the rules of conversational scorekeeping, which
allow for accommodations.
This means that in assessing a conversational record, the accommoda-
tional character of interpreting literal rules of dialectical procedure allow
316 INFORMAL FALLACIES

for a certain filling in of presuppositions and premisses, allowed to be made

true by the context of the conversation. However, as David Lewis points
out, a shift in the conversational context may also dictate a shift in the con­
versational scorekeeping, and with it a shift in the truth-values of state­
ments in the conversation.
It is because of the pragmatic character of conversational scorekeeping
that the use of games of dialogue to analyse actual conversational argumen­
tation can, with some justification, be called "informal logic." Otherwise,
the contention that there is essentially no difference between formal fal­
lacies and informal fallacies would be correct. It is a longstanding tradition
to divide fallacies into informal ones like those we have considered, and
formal ones. By "formal" ones are usually meant instances of invalid forms
of classical logic, e.g. 'If I pass, I have paid tuition. I have paid tuition,
therefore I pass.' This argument is an instance of the invalid form 'p ⊃ q, q,
therefore p.'
According to the analyses given in this book, the traditional fallacies all
turn out to be modelled, at least to some extent, by underlying formal struc­
tures that determine which forms of criticisms or rebuttal are correct or
incorrect. Are we then committed to the consequence that all informal fal­
lacies are really formal fallacies after all? If so, there would be no informal
logic as a separate discipline, but only difficulties in applying formal logics.
Perhaps these difficulties could just be due to the lack of systematic
development of many of the appropriate formal structures, a formalist
might insist.
While there is some truth in this line of argument, it fails to be entirely
correct or insightful because of the pragmatic gaps of applying formal sys­
tems to actual argumentation which need to be filled in by accommodating
principles of conversational scorekeeping. This means that the adjective
"informal" is partly inappropriate. Where it becomes inappropriate is in the
suggestions that formal structures have no role to play at all in the evalua­
tion of argumentation. Because of this unwelcome interpretation that erects
an unproductive barrier between formal logic and informal argumentation,
alternative names like "applied logic," "practical logic" or "the logic of
argumentation" would be preferable. However, it is because conversational
scorekeeping always outruns precisely regulated formal or dialectical rules
that a pragmatic element is desirable, and the term "informal logic" is not
altogether inappropriate.
Another reason why the term "informal logic" has a certain approp-
 INFORMAL LOGIC AS A DISCIPLINE 317

riateness is that the job of argument analysis that characterizes the sort of
work that the texts in this field of endeavor attempt involves a creative ele­
ment. For the task typically attempted is to confront a corpus of given argu­
ment (a case) that is incomplete, that has many gaps and uncertainties of
interpretation. To a certain extent, the argument critic must "invent" argu­
ments, or parts of them, to fill these gaps. Our thesis has been that the critic
should not take too many liberties in his "inventions," that he should oper­
ate under reasonable constraints.
We should make a careful distinction here between construction, or
invention of an argument, and evaluation of a given argument. Thinking of
a new line of argument could be called a process of invention. Of course, in
realistic arguments in natural language the process of argument pro and con
is never or rarely closed off altogether. It is always possible to think of a
new criticism, argument, or rebuttal — or to modify your position in
response to criticism.
In recent times however, logic has usually been taken to refer to the
job of evaluating a given argument. This traditional use of formal logic to
determine validity or invalidity starts from a given set of (designated) pre­
misses. The procedure, usually conceived, presumes that the argument is
already constructed and "closed off." This presumption met, the evaluator
checks for validity or other structural properties of the given set of proposi­
tions.
But should this be all there is to the reasoned analysis of arguments?
Our account of argument analysis has indicated that we can, and should in
some cases, go beyond that point. This job involves adding in more steps or
missing parts of the argument where such moves can be justified by the
appropriate rules of dialectic. Such moves, as we have studied them, should
not be untrammelled or gratuitous inventions. Informal logic is the set of
rules and structures that justifies the reasoned use of this process of
analysis.
There are certain historical antecedents for this "new" applied
approach to argument analysis. Indeed, the sort of approach I have been
arguing for is by no means altogether new.
The quest for a philosophical basis for the kind of subject that in recent
years has been suggested by the demand for an informal logic or method of
argument analysis, in my view, takes us back to the model of argument cal­
led 'dialectical' by Aristotle. According to Aristotle {Topics 101 a37),
dialectic is a kind of argument that can be used to discuss the status of
318 INFORMAL FALLACIES

reasoning and axioms in science because the aim of dialectic is to work

towards agreement by a question-answer process in a disputation. In dialec­
tic, the questioner does not require in every instance a premiss that is better
established or known than the conclusion he needs to prove. What he needs
are premisses that the answerer will concede. And from these commit­
ments, his aim is to prove his thesis. By contrast, in a demonstrative as
opposed to a dialectical argument, the prover must always work from pre­
misses better established than the conclusion he sets out to prove.
According to Aristotle, a characteristic of dialectic as a type of argu­
ment is that it can argue from premisses that are plausible, even if they fall
short of being known to be true on the basis of firm evidence.
The idea that dialectic might have an aspect that relates to the inven­
tion of arguments is one that is suggested by the history of the so-called
topics in logic.
In Aristotle, the concept of a topic (topos) was a device that enabled an
arguer to confront an argument and respond to it effectively by constructing
new arguments for or against it. Aristotle's list of the topics depended on
his concept of the predicables: genus, species, definition, differentia, prop­
erty and accident. Hence a topic is a kind of strategy of argumentation that
enables one engaged in dialectic to construct new arguments for or against
a position.
So construed, the topic seems to function as a creative device that ena­
bles one to discover or invent new arguments. Cicero and Boethius thought
of the topics in this way. For them, dialectic is the art of finding arguments,
as opposed to logic, the method of judging arguments — see Stump (1978,
p.25).
However, according to Stump, there was a tendency in the medievals
after Boethius to absorb the study of topics into the method of judging the
validity of an argument.
Bird (1960) shows that Abelard used the topic as an inference rule that
helps an argument evaluator find what is missing (enthymematic proposi­
tions) in an inference. The topic provides a relationship that makes an
imperfect syllogism valid by adding premisses. For example, according to
Abelard, the inference 'If it is man, it is animal' is justified as a good infer­
ence by the Topical Maxim of Species: of whatever the species is predi­
cated, so is the genus (Bird, 1960, p. 144). It is also interesting to note that
Abelard recognized less perfect enthymemes in which there is no necessity
of consecution. An example given by Bird (1960, p.142) is that it might be
 INFORMAL LOGIC AS A DISCIPLINE 319

argued from the fact that I ran off with this girl that I love her.
The next historical development took place with William of Ockham,
where the study of topics became absorbed into the theory of conditionals.
These later developments tended to think of the topics less and less a
technique for inventing new arguments. And hence the study of topics
tended to become absorbed into the dominant conception of logic as a
method for judging the validity of arguments.
Perhaps then the history of logic is repeating itself to some extent. The
basic problem of the new informal logic is for us to see how the reasonable
critic or evaluator of an argument can reasonably fill in enough of the
missing bits to give a significant and helpful evaluation without himself tak­
ing the step of entering into the argument as a participant. For that would
be too much "creativity" — beyond the objectivity and distance required of
the critic.
It seems to me the would-be informal logician at present is characteris­
tically put in the position of one who must evaluate an argument where the
argument is partly hidden or missing. Not all the relevant premisses and
conclusions are given to him, just some of them. Therefore, he must do his
best to fill in the missing parts in a fair and justifiable way, before he can go
on to apply standards which will evaluate the argument as open or closed to
the types of criticisms of arguments that are made in informal logic. Hence
his job, in the initial stages, is somewhat like that of an archeologist of argu­
ment. He must look over the existing bits of argument he is given — some­
what like the archaeologist looks at ruins or fossils — and then arrive at
some reasoned judgement or reconstruction of what the whole entity might
have looked like, judging by the existing traces.
The two basic tools the analyst of argument has, according to the
theory of argument expounded in the foregoing account, is the structure of
a game of dialogue and the structure of the graph of an argument. By this
account, the informal logician is directed to the task of attempting to recon­
struct the context of dialogue — the arguer's position, his ultimate thesis,
his strategy and the questions asked and answered according to rules of
dialogue that evolved his argument. Then he has the job of mapping out the
graph of the overall network of argumentation as juxtaposed against the
context of dialogue. This means filling in many presumed premisses and
conclusions in a way that fairly represents the position of the arguer whose
argument is being evaluated or criticized. It also involves accounting for the
opposing side of the argument in a fair and reasonable way. From this
320 INFORMAL FALLACIES

reconstruction, there can evolve fair and justifiable criticisms of an argu­

ment, as well as replies or rebuttals of those criticisms. The informal logi­
cian's job is to arrive, on reasonably well-argued evidence, at an evaluation
of these criticisms. Even where the analysis does not decisively refute or
establish the target argument, it can still be valuable and informative.

7. Pedagogical Directions for Informal Logic

The proliferation of "new wave" textbooks of informal logic suggests
the beginnings of a new subject. There is a very genuine need for this sub­
ject, and some of the new texts, especially in the past two years, show an
increasing practical sophistication in the use of applied argument analysis
techniques. The main problem continues to be the shortage of serious
investigation of the foundations of this new field. Clearly, the textbooks are
whistling in the dark. Authors of textbooks have even confided to me that
they feel a little guilty teaching in this area because, being constantly forced
to oversimplify complex matters in order to make their points, they wonder
whether this "lying" is justified. Even if it is, the feeling of possible subter­
fuge can make one uneasy.
Perhaps the superficiality of many textbooks in this area is a reflection
that the motivation of informal logic courses has been to expand enrolment
by finding a subject-matter that will appeal to, or be tolerated by, less able
or less motivated students. Even so, the more serious students will only tol­
erate the Standard Treatment for so long, and then some will begin to ear­
nestly inquire whether there really are objective guidelines or procedures to
effect an orderly classification of arguments into "correct" or "incorrect"
bits of reasoning. If there is no good generally applicable answer why a par­
ticular example labelled an ad hominem or ad verecundiam fallacy is falla­
cious or not, by appeal to some well-formulated rule or general procedure,
the students are rightly entitled to conclude that the course is going
nowhere.
Sometimes the remedy used is to introduce a stiff dose of classical
logic, but if that goes on long enough the students will correctly perceive
that it is not related in any way — as these courses are often taught, at any
rate — to the fallacies. Thus what is needed now in informal logic is a rap­
prochement of formal procedures and informal fallacies and arguments,
and a deepening of the methodological roots of informal logic by the study
of the structures of arguments. My own suggestion is that these two direc­
tions are one and the same at bottom.
 INFORMAL LOGIC AS A DISCIPLINE 321

Fortunately, some of the texts that have recently begun to appear, as

well as some of the older texts, do successfully bring some structure and
sophistication to bear on the treatment of the fallacies and related problems
of argument analysis. Moreover, as we have seen, there are resources of a
theoretical nature available from some quarters that can be marshalled to
give some coherence to the collection of rudimentary techniques now being
used by these texts. Moreover sources like the Informal Logic Newsletter
provide new resources for serious investigation of this area as a subject for
scholarly research. Part of the problem is getting enough teachers of
philosophy to take seriously the claim that there is a genuine area here for
legitimate scholarly research in its own right.
However, I believe that such a definite need has always been implicit
in the claims of logic as taught within philosophy departments. Our man­
date is not simply to investigate formal structures for their purely mathe­
matical interest, but to teach logic as a criterion for correct argument that is
applicable to arguments in domains other than the purely mathematical.
The legitimate attraction of a logic course taken in the philosophy depart­
ment is that is should help the student to reason more clearly or soundly in
areas as diverse as public affairs, the consumer marketplace, ethics, or any­
where where real arguments are advanced and criticized. I do not believe
that this is exclusively the domain of English composition or of pure
mathematics. There is surely a broad area of "informal logic" somewhere in
between that is and was supposed to be taught in philosophy departments.
I think we are just not doing a very good job of working on applied logic as
a discipline worthy of study in its own right at the moment, and need to pull
up our socks. If we are going to teach the subject, we should try to find out
what it is all about. And in fact we are now teaching it on a wide scale in
departments of philosophy, in my opinion.
The recommendations I make are the following. First, informal logic
should broaden its pedagogical scope of study to include not only the detec­
tion of fallacies in existing arguments in newspaper articles, and other writ­
ten sources or reports. More broadly, the discipline should adopt as a
pedagogical methodology the study of actual case disputes on particular
questions in topic. First, the arguments should be set out, pro and con, in a
case study on a particular issue or dispute. Once the thesis of each position
is formulated, rebuttals and objections should be constructed, and further
replies considered. Thus conceived, the study of argument takes criticism on
as one important aspect, but also studies reasoning by disputation as a con-
322 INFORMAL FALLACIES

structive method of inquiry in its own right.

I have argued that formal dialectic is the most appropriate theoretical
model for rational disputation so conceived. Therefore, my second recom­
mendation is that the theoretical roots of disputational argument be
deepened by the study of formal games of dialogue and the applicable mod­
els of inference, be they classical or non-classical logics and structures, con­
tained within formal games. The area now widely recognized as pragmatics
by linguists needs to be taken more seriously by all who profess an interest
in or commitment to informal logic.
These recommendations represent the best future directions for infor­
mal logic as a discipline if it is to fulfill the needs and objectives it appears
to have set for itself, given current pedagogical practices.
The game of dialogue as a model of argument allows for the develop­
ment of discriminations and refinements in arguments by its contestive and
developing nature as a dynamic interaction of participants. It forces the
participant to look to the evidential roots of his thesis to try to defend it,
and to react critically to the arguments of the opponent. Needless to say,
such a shift in our perception of logic as a discipline toward the dialectical,
and toward the applied case study of actual disputations, represents a sig­
nificant departure from many current pedagogical and scholarly practices
and traditions.

NOTES

1. See the Informal Logic Newsletter, iv. no.2, 1982, p.l [Editor's note].
2. See R.H. Johnson and J.A. Blair, 'The Recent Development of Informal Logic,' in R.H.
Johnson and J.A. Blair, eds., Informal Logic: The First International Symposium, Inverness,
California, Edgepress, 1980, 3-28.
3. Scriven (1976) is an exponent of this view.
4. Ilid.,.38.
5. See Tarski (1956) on the liar paradox and related semantical paradoxes.
6. See Scriven (1976).
7. See chapter six, section seven.
 BIBLIOGRAPHY

Anscombe, Gertrude Elizabeth

1957 Intention. Oxford: Blackwell.

Apostel, Leo
1982 "Towards a general theory of argumentation". In Barth and Martens (eds.) 1982, 93-
122.

Åqvist, Lennart
1974 "A new approach to the logical theory of actions and causality". In Soren Stenlund
(ed.) 1974, Logical theory and semantics. Dordrecht: Reidel, 73-91.

Aristotle
1958 Topica et sophistici elenchi. Translated by W.A. Pickard-Cambridge, ed. by W.D.
Ross. Oxford: The Clarendon Press.

Anderson, Alan R., and Nuel D. Belnap, Jr.

1975 Entailment: The logic of relevance and necessity, Vol.1. Princeton, N.J.: Princeton
University Press.

Bailey, Frederick G.
1983 The tactical uses of passion. Ithaca and London: Cornell University Press.

Barth, Else, and Erik C.W. Krabbe

1982 From axiom to dialogue. New York: De Gruyter.

Barth, Else, and J.L. Martens

1977 "Argumentum Ad Hominem: From chaos to formal dialectic". Logique et Analyse
76-96.

Barth, Else, and J.L. Martens, eds.

1982 Argumentation: Approaches to theory formation. Amsterdam: John Benjamins.

Beardsley, Monroe C.
1950 Practical logic. New York: Prentice-Hall.

Belnap, Nuel D., Jr.

1963 An analysis of questions: Preliminary report. Santa Monica: System Development
Corporation.

Bird, Otto
1960 "The formalizing of the topics in mediaeval logic." Notre Dame Journal of Formal
Logic 1.138-149.
324 INFORMAL FALLACIES

Burge, Tyler
1977 "A theory of aggregates." Noûs 11.97-117.

Campbell, Stephen K.
1974 Flaws and fallacies in statistical thinking. Englewood Cliffs: Prentice-Hall.

Carlone, Frank C , Karol Dycha, and Leo Raffin

1981 An informal logic workbook. Windsor: Kate Parr.

Carlson, Lauri
1983 Dialogue-games. Dordrecht: Reidel.

Cederblom, Jerry, and David W. Paulsen

1982 Critical reasoning. Belmont, California: Wadsworth.

Cohen, David
1973 The crucial 10% that really counts for trial victories. Englewood Cliffs, N.J.: Execu­
tive Reports Corporation.

Copi, Irving M.
1972 Introduction to logic, 4th ed.. New York: MacMillan.

Crossley, David J., and Peter A. Wilson

1979 How to argue. New York: Random House.

DeMorgan, Augustus
1847 Formal logic. London: Taylor and Walton.

Ellul, Jacques
1972 Propaganda. New York: Knopf.

Engel, S. Morris
1976 With good reason: An introduction to informal fallacies. New York: St. Martin's
Press.

Epstein, Richard L.
1979 "Relatedness and implication." Philosophical Studies 36.137-173.

1981 Relatedness and dependence in propositional logics. Unpublished manuscript.

Finocchiaro, Maurice A.
1974 "The concept of ad hominem argument in Galileo and Locke." The Philosophical
Forum, 5.394-404.

1980 Galileo and the art of reasoning. Dordrecht: Reidel.

1981 "Fallacies and the evaluation of reasoning." American Philosophical Quarterly

18.13-22.

Fogelin, Robert J.
1978 Understanding arguments. New York: Harcourt Brace Jovanovich.

Govier, Trudy
1981 "Worries about Tu Quoque as a Fallacy." Informal Logic Newsletter 3:3.2-4.
 BIBLIOGRAPHY 325

1983 "Ad Hominem: Revising the textbooks." Teaching Philosophy 6.13-24.

Green, Romuald
1963 An introduction to the logical treatise 'De Obligationibus' with critical texts of Wil­
liam of Sherwood and Walter Burley. Thesis, Catholic University of Louvain.

Grice, H. Paul
1975 "Logic and conversation." In D. Davidson and G. Harman (eds.) 1975, The Logic
of Grammar. Encino, Calif.: Dickenson Publishing Co., 64-75.

Groarke, Leo
1982 "When two wrongs make a right." Informal Logic Newsletter 5:1.10-13.

Hamblin, Charles L.
1970 Fallacies. London: Methuen.

1971 "Mathematical models of dialogue." Theoria 37.130-155.

Harper, Charles W., Jr.

1980 "Relative age inference in paleontology." Lethaia 13.239-248.
Harrah, David
1963 Communication: A logical model. Cambridge, Mass.: M.I.T. Press.

1980 "On speech acts and their logic." Pacific Philosophical Quarterly 61.204-211.

1984 "Message theory and the semantics of dialogue". In Lucia Vaina and Jaakko Hin-
tikka (eds.) Cognitive constraints on communication. Dordrecht: Reidel, 267-276.

Hegselmann, Rainer
1982 Formale Dialektik: Ein Beitrag zu einer Theorie des Rationalen Argumentierens.
Unpublished manuscript.

Hintikka, Jaakko
1976 The semantics of questions and the questions of semantics. Amsterdam: North-Hol­
land.

1981 "The logic of information-seeking dialogues: A model." Erkenntnis 38, 1979, 355-
368. Reprinted in Werner Becker and Wilhelm . Essler (eds.) 1981, Konzepte der
Dialektik. Frankfurt am Main: Vittorio Klostermann, 212-231.

Hintikka, Jaakko, and Esa Saarinen

1979 "Information-seeking dialogues: Some of their logical properties." Studia Logica
38.355-363.

Hull, David L.
1967 "Circularity and certainty in evolutionary taxonomy." Evolution 21.174-189.

Iseminger, Gary
1980 "Is relevance necessary for validity?" Mind 89.196-213.

Johnson, Ralph H., and J. Anthony Blair

1977 Logical self-defense. Toronto and New York: McGraw-Hill Ryerson.
326 INFORMAL FALLACIES

1980 Informal logic: The first international symposium. Inverness, California: Edgepress.

Johnstone, Henry W., Jr.

1970 "Philosophy and Argumentum Ad Hominem revisited." Revue Internationale de
Philosophie 24.107-116.

1978 Validity and rhetoric in philosophical argument. University Park, Pennsylvania:

Dialogue Press of Man and World.

Kahane, Howard
1978 Logic and Philosophy. Belmont, California: Wadsworth.

Kamiah, Wilhelm, and Paul Lorenzen

1973 Logische Propädeutik. Mannheim: Hochschultaschenbücher-Verlag.

Kasher, Asa
1979 "What is a theory of use?" In Avishai Margalit (ed.) 1979, Meaning and Use. Dor­
drecht: Reidel, 37-55.

Kielkopf, Charles
1980 "Relevant appeals to force, pity, and popular pieties." Informal Logic Newsletter
2:2.1-5.

King, John L.
1979 "Bivalence and the law of excluded middle." American Philosophical Quarterly
16.17-25.

Kornhauser, Ruth
1978 Social sources of delinquency. An appraisal of analytic models. Chicago and Lon­
don: University of Chicago Press.

Krabbe, Erik C.W.

1982 "Theory of argumentation and the dialectical garb of formal logic." In Barth and
Martens (eds.) 1982, 123-132.

1982 Studies in dialectical logic. Doctoral dissertation at Rijksuniversiteit te Groningen.

Lehnert, Wendy
1984 "Problems in question answering." In Lucia Vaina and Jaakko Hintikka (eds.)
1984, Cognitive Constraints on Communication. Dordrecht: Reidel 137-159.

Lenk, Hans
1976 "Handlungsgraphen: Graphentheoretische Modelle in der Analytischen Hand­
lungsphilosophie." Grazer Philosophische Studien 2.159-172.

Lewis, David
1979 "Scorekeeping in a language game." In R. Bauerle, U. Egli, and A. von Stechow
(eds.) 1979, Semantics from different points of view. Heidelberg and New York:
Springer-Verlag, 172-187.

1982 "Logic for equivocators." Noûs 16.431-441.

Lorenzen, Paul
1969 Normative logic and ethics. Mannheim: Hochschultaschenbücher-Verlag.
 BIBLIOGRAPHY 327

Lukasiewicz, Jan
1957 Aristotle's syllogistic from the standpoint of modern formal logic, 2nd edition. Lon­
don: Oxford University Press.

Mackenzie, James D.
1979 "How to stop talking to tortoises." Notre Dame Journal of Formal Logic 20.705-
717.

1981 "The dialectics of logic." Logique et Analyse 94.159-177.

Makinson, David
1973 Topics in modern logic. London: Methuen.

Manor, Ruth
1979 "A language for questions and answers." Theoretical Linguistics 6.1-21.

1981 "Dialogues and the logics of questions and of answers." Linguistische Berichte 73.1-
28.

1982 "Pragmatics and the logic of questions and assertions." Philosophica 29.45-95.

1983 Representing dialogues. Manuscript, Universität Konstanz.

Massey, Gerald J.
1981 "The fallacy behind fallacies." Midwest Studies in Philosophy 6.489-500.

McPeck, John E.
1981 Critical thinking and education. Oxford: Martin Robertson & Co.

Mitroff, Ian I., and Richard O. Mason

1981 Creating a dialectical social science: Concepts, methods, and models. Dordrecht:
Reidel.

Morgan, Edmund M.
1963 Basic problems of evidence, 4th ed.. Philadelphia: American Law Institute.

Parry, William T.
1933 "Ein Axiomsystem für eine neue Art von Implikation (Analytische Implikation)".
Ergebnisse eines Mathematischen Kolloquium 4.5-6.

Perelman, Chaim
1982 The realm of rhetoric. Notre Dame: University of Notre Dame Press.

Priest, Graham
1979 "The logic of paradox." Journal of Philosophical Logic 8.219-241.

1980 "Sense, entailment and modus ponens." Journal of Philosophical Logic 9.415-435.

Rastall, Robert.H.
1956 "Geology." Encyclopaedia Britannica 10.168.
Rescher, Nicholas
1976 Plausible reasoning. Assen-Amsterdam: Van Gorcum.
328 INFORMAL FALLACIES

1977 Dialectics. Albany: State University of New York Press.

Salmon, Wesley
1963 Logic. Englewood Cliffs, N.J.: Prentice-Hall.

Schank, Roger
1984 "Looking for a process model of dialogue." In Lucia Vaina and Jaakko Hintikka
(eds.) 1984, Cognitive constraints on communication. Dordrecht: Reidel, 161-173.

Scriven, Michael
1976 Reasoning. New York: McGraw-Hill.

Schwartz, Louis E.
1973 Proof, persuasion and cross-examination, vol. 2. Englewood Cliffs, N.J.: Executive
Reports Corporation.

Shoesmith, David J., and Timothy J. Smiley

1980 Multiple-conclusion logic. New York: Cambridge University Press.

Sidgwick, Alfred
1884 Fallacies. New York: Appleton.

Slupecki, Jerzy, Grzegorz Bryll, and Urzula Wybraniec-Skardowska

1971 "Theory of rejected propositions I." Studia Logica 29.75-123.

Stump, Eleonore
1978 Boethius's de topicis diferentiis. Trans, with notes and essays. Ithaca and London:
Cornell University Press.

Tarski, Alfred
1956 Logic, semantics, metamathematics. Trans. J.H. Woodger. Oxford: Oxford Univer­
sity Press.

Vanderveken, Daniel
1980 "Illocutionary logic and self-defeating speech acts." In John R. Searle, Ferenc
Kiefer, and Manfred Bierwisch (eds.) 1980, Speech Act Theory and Pragmatics.
Dordrecht: Reidel, 247-272.

Van Dun, Frank

1972 "On the modes of opposition in the formal dialogues of P. Lorenzen." Logique et
Analyse 15 (Ν.S.).103-136.

1982 "On the philosophy of argument and the logic of common morality." In Barth and
Martens (eds.) 1982, 281-294.

Van Eemeren, Frans H., and Rob Grootendorst

1983 Speech acts in argumentative discussions. Dordrecht: Foris Publications.

Van Eemeren, Frans H., Rob Grootendorst, and T. Kruiger

1984 The study of argumentation. New York: Irvington Publishers.

Von Wright, Georg H.

1963 "Practical inference." The Philosophical Review 72.159-179. Reprinted in von
Wright 1983, 1-17.
 BIBLIOGRAPHY 329

1968 An essay in deontic logic and the general theory of action. Amsterdam: North-Hol­
land.

1972 "On so-called practical inference." Acta Sociologica 15.39-53. Reprinted in von
Wright 1983, 18-34.

1983 Practical Reason. Ithaca: Cornell University Press.

Walton, Douglas N.
1979 "Philosophical basis of relatedness logic." Philosophical Studies 36.115-136.

1980 "Why is the Ad Populum a fallacy?" Philosophy and Rhetoric 13.264-278.

1980 "On the logical form of some commonplace action expressions." Grazer
Philosophische Studien 10.141-148.

1980 "Ignoratio Elenchi: The red herring fallacy" Informal Logic Newsletter 2:3.3-7.

1980 "Petitio Principii and argument analysis." In R.H. Johnson and J.A. Blair (eds.)
1980, Informal Logic. Pt. Reyes, California: Edgepress, 41-54.

1981 "The fallacy of many questions." Logique et Analyse 95/96.291-313.

1982 Topical relevance in argumentation. Amsterdam: John Benjamins.

1983 "Enthymemes." Logique et Analyse 103/104.395-410.

1984 Logical dialogue-games and fallacies. Lanham, Maryland: University Press of

America.

1985 Arguer's position. Westport: Connecticut: Greenwood Press.

Walton, Douglas N., and Lynn M. Batten

1984 "Games, graphs and circular arguments." Logique et Analyse 106.133-164.

Woods, John, and Douglas Walton

1974 "Argumentum Ad Verecundiam." Philosophy and Rhetoric 7.135-153.

1975 "Petitio Principii." Synthese 31.107-127.

1976 "Ad Baculum." Grazer Philosophische Studien 2.133-140.

1976 "Fallaciousness without invalidity?" Philosophy and Rhetoric 9.52-54.

1977 "Ad Hominem contra Gerber." Personalist 58.141-144.

1977 "Ad Hominem." The Philosophical Forum 8.1-20.

1977 "Composition and division." Studia Logica 36.381-406.

1977 "Post Hoc Ergo Propter Hoc." Review of Metaphysics 30.569-593.

1977 "Petitio and relevant many-premissed arguments." Logique et Analyse 77/78.97-
110.
330 INFORMAL FALLACIES

1978 "Arresting circles in formal dialogues." Journal of Philosophical Logic 7.

1978 "The fallacy of Ad Ignorantiam." Dialectica 32.87-99.

1979 "Equivocation and practical logic." Ratio 21.31-43.

1979 "Circular demonstration and von Wright-Geach entailment." Notre Dame Journal
of Formal Logic 20.768-772.

1979 "Laws of thought and epistemic proofs." Idealistic Studies 9.55-65.

1982 Argument: The logic of the fallacies. Toronto and New York: McGraw-Hill Ryer-
son.

1982a "Question-begging and cumulativeness in dialectical games." Noûs 4.585-605.

1982b "The Petitio: Aristotle's five ways." Canadian Journal of Philosophy 12.77-100.

Whately, Richard
1836 Elements of logic. New York: William Jackson.
Zadeh, Lofti A.
1975 "Calculus of fuzzy restrictions." In Aldo de Lucca (ed.) 1975, Fuzzy sets and their
applications to cognitive and decision processes. New York: Academic Press, 1-39.
 INDEX OF NAMES

Abelard, Pierre, 318 Darwin, Charles, 308-09

Ad Hominem, The Philosophical DeMorgan, Augustus, 61, 186, 223, 224
Forum, 240 De Sophisticis Elenchis, 214
Anderson, Alan, R., 262, 273, 276
Anscombe, Gertrude Elizabeth, 40 Ellul, Jacques, 61
Aqvist, Lennart, 99 Empiricus, Sextus, 303
Aristotle Engel, S. Morris, 8
dialectic, 98-99, 317-18 Epstein, Richard L., 83-84, 96, 301
enthymemes, 133, 135, 148 Essay in Deontic Logic and the General
fallacy of composition, 214 Theory of Action, An, 240
ignoratio elenchi, 10, 115-18
petitio principii, 170, 172, 175 Foster Parents Plan of Canada, 42
practical syllogism, 40 Friesen, Benno, 121-22

Bailey, Frederick G., 33, 38 Geach, Peter, 158-160

Batten, Lynn M., 161, 175 Goebbels, Joseph, 35
Beardsley, Monroe C., 158-60 Govier, Trudy, 240
Beauchesne, 121 Grice, H. Paul, 81
'Belief in the Law of Small Numbers' Grootendorst, Rob, 41, 134, 135-36
Psychological Bulletin 76 1971, 205
Belnap, Nuel D . , 99, 112, 114, 262, 273 Hamblin, Charles L., 4, 8
276 ad mìsericordiam, 42
Bird, Otto, 318 ad verecundiam, 188
Blair, J. Anthony, 46, 160, 322 amphiboly, 213
Blaker, Rod, 121 Aristotle, 10
Boethius, Anicius Manlius Severinus, circular arguments, 158
318 complex arguments, 158
Broadbent, Edward, 11-15 dialogue games, 99, 100, 101, 125,
Burge, Tyler, 215 137, 294, 297, 299, 301
equivocation, 242, 245, 248, 249-52,
Canadian Criminal Cases, 193 254, 258, 284
Carlone, Frank C , 160 Hamblin games, 99, 104,106,137-38,
Cederblom, Jerry, 5, 246, 248, 284 232, 293, 294, 297
Cicero, Marcus Tullius, 318 Hansard (Debates of the House of
Cohen, David, 196, 197 Commons of Canada)
Copi, Irving M., 17, 47, 117, 160 November 30, 217
Cresswell, Max, 131 September 22, 219-21
Crosbie, John, 219-22 February 17, 11-12
332 INDEX OF NAMES

March 6, 23 Outlines of Pyrrhonism, 303

March 12, 119-20, 121, 122-23
Harper, Charles W., Jr., 306, 306-08, Paulsen, David W., 5, 246, 248, 284
311 Perelman, Chaim, 33
Harrah, David, 99 Popper, Karl, 191
Hintikka, Jaakko, 99, 101, 107, 114-15, Priest, Graham, 261-66, 266-67, 268,
125, 294 272, 282-83, 285, 287
Hintikka games, 101, 102, 104, 107, Prior Analytics, 170
125, 140-41,293-94
Hnatyshyn, Ray, 119-21 Quine, Willard van Orman, 242
Hull, David L., 309-11
Rae, Bob, 52-60
Informal Logic: The First International Rastall, Robert H., 305-07
Symposium, 322 Rescher, Nicholas, 173-74, 189, 294
Informal Logic Newsletter Vol. 4 no. 2, Reuther, Walter, 38
322 Russell, Bertrand, 233
Informal Logic Workbook, 57-58
Salmon, Wesley, 189, 203
Johnson, Ralph, H . , 46, 160, 322 Schwartz, Louis E., 226
Johnston, Donald, 23, 23-25, 220-22 Schwemmer, Oswald, 102
Scriven, Michael, 160
Kahnman, Daniel, 205 Shoesmìth, David J., 160
Kalman, 274 Siddon, Thomas, 23-25
Kamlah, Wilhelm, 201 Smiley, Timothy, 160
Kielkopf, Charles, 44 Stump, Eleonore, 318
King, John L., 211
Kornhauser, Ruth, 311, 313 Tarski, Alfred, 291
Taylor, Gordon, 122-23
Lalonde, Marc, 119-21 Topics, The, \1
Lewis, David, 82, 259, 260, 261-62, 267- Trudeau, Pierre Eliot, 11-15
69, 270, 273, 279, 315-16 Tversky, Arnos, 205
Lorenz, Kuno, 102
Lorenzen, Paul, 102, 103, 294 van Eemeren, Frans H., 41, 134-36, 138
Lorenzen games, 102-04, 106-07, 293 von Wright, Georg H . , 40-41, 240

MacEachen, Allan J., 52-58 Whately, Richard, 61, 223

Mackenzie, James D . , 297 William of Ockham, 319
Makinson, David, 276, 278, 285 Woods, John, 179, 182, 183, 191, 207,
Manor, Ruth, 100 215, 240, 242, 243, 259, 311
Mason, Richard O., 195, 304 'Worries About Tu Quoque as a Fal­
McPeck, John E., 302 lacy' Informal Logic Newsletter Vol. 3,
Mill, John Stuart, 157 No. 3, 240

Newman, Cardinal, 218 Zadeh,Lofti Α., 211

 GENERAL INDEX

Affirming the consequent, 12, 13-14, 86 debates, in, 235

Amphiboly, 8, 241 dialectic, must be analyzed within,
Argument, 1-2 See also Circular argu­ 235-36
ment and Debate dialectical context, 233
conclusion, intermediate, 158 difficulty in pinning down, 235
convergent, 158 evaluating, check points for, 227-29
definition, 16-21 inconsistency, circumstantial allega­
definition, broad, 253 tion turns on, 219
deductive and inductive, identifying, inconsistency, as allegation of deontic
205-06 pragmatic, 236
dialectical, 27, 98-99 legal argument, in, 51
dialectical contrasted with reproduci­ poisoning the well, 218
ble, 26 reasonable, can be, 26, 230-31, 238
dialectical model, 3 relevance, as failure of, 218
divergent, 159 tu quoque, legitimate function of, 22-
evaluating objectively, 304 25
form of, 66-67, 68 Argumentum ad ignorantiam, 9, 105-06,
forms, can have different, 78 131
function, 27 Argumentum ad misericordiam, 5, 42,
graphing, 160-62, 178-80 44-45
indirect, 17 in legal argument
legal involves fallacies, 51 Argumentum ad populum, 5, 33-38, 45-
model and actual, gap between, 301 46
monolectical, 27 Argumentum ad verecundiam, 6, 185
serial, 159 ad ignorantiam, may be a form of,
validity, deductive, 63-65, 67 106
validity, testing for, 72-73 conditions of correctness, 185-87
validity with invalid form, 78, 79 type of argument, 188-89
ways of winning, 19-20
Argumentum ad baculum, 5, 39, 41 Biased statistics, 203, 204-05
Argumentum ad hominem abusive, 47,
217 Circular argument, 157, 170, 171, 180-
attack, 239 82
attack distinguished from refutation, see also petitio principii banned by
239 plausible condition, 174
circumstantial, 6, 47, 48, 217 fallacious, need not be, 173, 176
circumstantial, four types, 230-33 graphs to detect, 179, 180
confusion, based on, 27 Commitments, dark-side, 125, 142
334 GENERAL INDEX

Commitment-store, 100, 137 283-84, 286

Complex questions see Fallacy of many analysis requires broad definition of
questions and Questions argument, 253-54
Conversational theory of argument, 81 contextual shift, depends on, 242-43
criticism provisional, 248
Debate, 48-51 description, 241-42
ad hominem in, 235 Hamblin's analysis, 249-53, 254, 258
fallacies, is open to, 61 logical games susceptible, 244, 267,
illogicality, 60 272, 282-83, 285
parliamentary, 50 many arguments, as, 257
Delphi technique, 188 natural language, depends on, 258-59
Denying the antecedent, 73, 74 possibility cannot be excluded, 286
Dialectic, 130, 295-96 RM, in, 276-83
Aristotle's conception, 317-18 strategies for dealing with, 284
role of formal logic in, 295 strategy in dialogue game, as, 251-52,
Dialectical argument, 27, 98 253
Dialogue games, 90-91 ways of combating, 244-45
conventionalism tolerated in, 92 Expert testimony, 192-93 see also
definition, 21, 97 argumentum ad verecundiam
flexibility of, 295-96 challenged by lay person, 198
informality of, 316 Fallacies, see also particular names
strategy in, 145-46 classification of, 9
strictness of, 293-94 empirical sciences, study linked to, 94
types, 127-29 fallacious, are not always, 3, 94
ABV, 126-29, 296 informal contrasted with formal, 316
CBV, 296, 304-06 rule violation, as, 95
Hamblin games, 99, 104, 106,137-38, valid, must appear, 94
231,293,294,297 Fallacy of composition, 8, 214, 215
Hintikka games, 101, 102, 104, 107, Fallacy of division, 8, 214, 215
125, 140-41, 293, 294 Fallacy of many questions, 9, 110-11,
Lorenzen games, 102-04, 106-07, 293 118-19, 131 see also Questions
obligation game, 101 Fallacy of popularity, 45-46 see also
why-because-game-with-questions, Argumentum ad populum
105 Force, appeal to, see Argumentum ad
why-questions in, 297-301 baculum
Dilemma, 72
Gambler's fallacy, 204, 205
Enthymeme, 2, 133, 148, 149, 150, 154- Games of dialogue see Dialogue games
55,318 Graphs of arguments, 160-162
Aristotle's definition, 133 help to detect circles, 179, 180
determining, 133-36
in CBV, 146-48 purpose, 178-80
in Hamblin games, 138-39
in Hintikka games, 140-41 Hypothetical chain, 66, 67, 68, 71
Equivocation, 6-7, 212-13, 248, 259,
Ignoratio elenchi, 10-15, 115-16, 117
 GENERAL INDEX 335

Inconsistency Obligation game, 101

ad hominem allegation turns on, 219
appearance of, 230-32 Parliamentary debate, objective, 118-19
deontic-pragmatic, 236-37 Personal attack see Argumentum ad
effect of, 231-32 hominem
logical stronger than pragmatic, 232 Petitio principii, 9-10, 170, 175-80, 182-
positional, 237-38 83 see also Circular argument
practical, 225 applicability to particular subjects,
Inferences, paradoxical, 80 313
Innuendo, 234 see also Argumentum ad dependency conception, 182
hominem equivalency conception, 182
Insufficient statistics, 203, 204 Pity, appeal to see Argumentum ad mis-
Irrelevance see Ignoratio elenchi, ericordiam
Relatedness and Relevance Plausibility, 189
arguing from greater to lesser, 171,
Logic, formal, see also Propositional 172-73
calculus, Relatedness logic and Rele­ conditions on, 172-75
vance logic Plausible inference, 173, 189
ambiguity, for dealing with, 260 Position
applying, 73-74 defining, 236-37
dialectic, role in, 295 inconsistency of, 238
fallacies, applicability to, 292 Post hoc ergo propter hoc, 11, 206-07
fuzzy, 211 types, 206-09
inferences, paradoxical, 80, 83 Practical inference, 40, 224-25
informal, compared with, 316 necessary condition schema, 40
natural language, relation to, 291-92 sufficient condition schema, 40-41
uses, 65, 75-76 Pragmatics, 89, 89-90, 295
validity, deductive, 63-65, 79 Premisses, see also Enthymeme, 28,
validity, never a reliable test for, 79 170-71
value, 75 Proof compared with refutation, 28
Logic, informal application to particular Propositional calculus, classical, 68, 69,
subjects, 302, 303 71 see also Logic, formal argument
appropriateness of term, 294-95 forms
formal, compared with, 316 dilemma, 72
method of argument analysis, 317 hypothetical chain, 66, 67, 68, 71
philosophical basis, 317 modus ponens, 66, 68, 71, 72, 86
teaching, 320-21 modus tollens, 66, 71
usefulness, 289 constants
conjunction, 68
Modus ponens, 66, 68, 72, 86 disjunction, 68-69
Modus tollens, 66, 71 material conditional, 70-71
Misconception of refutation see negation, 68
Ignoratio elenchi fallacies
Monolectical argument, 27, 28 affirming the consequent, 72, 73-
74, 86-87
336 GENERAL INDEX

denying the antecedent, 73, 74 arguments which fail in, 265

relatedness logic is subsystem of, 86 conditional, 266
validity, test for, 72-73 conjunction in, 263-65
deducibility relation, 261
Quarrels, vagueness in, 18 disjunction in, 266
Questions see also Fallacy of many equivocation, application to, 267,
questions answer, direct, 113-14, 115, 282-83
124 equivocation, does not stop, 272,
Hintikka games, in, 125 284-85
presuppositions, 115, 125, 130 negation in, 262-63
presuppositions, multiple, 108-10, 111 validity in, 261, 268
presuppositions, unwelcome, 112 RM, 261-66
responses, 110 arguments invalid in, 273
safe, 112 axioms, 273
unfair, 29 conjunction, disjunction and nega­
whether-questions, 114 tion, 273
why-questions in dialogue games, disjunctive syllogism, 282
296-301 equivocation, allows, 282
equivocation, application to, 278-
Refutation compared with proof, 28 83
Relatedness logic, 83 implication, 274-75, 276, 285
affirming the consequent invalid in, modus ponens, 276-77
86
arguments which fail in, 86-88 Semantics, 88-90, 294
classical logic, relation to, 92 Shroud of Turin, 190-91
classical logic, is subsystem of, 88 Slippery slope argument, 7-8, 209-10,
constants 314
conditional, 83 Sorites argument, 210-11
conjunction, 85 Sportman's rejoinder, 47, 222-24, 228-
disjunction, 85 30
negation, 85 Statistical fallacies, 203-04
modus ponens valid in, 86 Strawman fallacy, 10
relations between simple and com­ Survival of the truth hypothesis, 50
plex propositions, 85-86
Relatedness of subject matters, 82-83 Threat, 42 see also Argumentum ad
Relationships, emotional, 33 baculum compared with warning
Relevance, see also Ignoratio elenchi
and Relatedness Vagueness, 248 see also Slippery slope
ignored in classical logic, 81 argument
pragmatic, 91 tolerance of, 314
semantic, 91 Validity, deductive, 63-65, 67
Relevance logics in system LP, 261, 268
LP, 260-66 test for, 72-73
with invalid form, 78-79